the political economy of trade, aid and foreign investment ... economy the... · the political...
TRANSCRIPT
THE POLITICAL ECONOMY OF TRADE,AID AND FOREIGN INVESTMENT POLICIES
Editors:
DEVASHISH MITRA
Department of Economics,
The Maxwell School of Citizenship & Public Affairs,
Eggers Hall, Syracuse University,
Syracuse, NY 13244, USA
ARVIND PANAGARIYA
School of International & Public Affairs,
International Affairs Building,
420 West 118th Street, Columbia University,
New York, NY 10027, USA
2004
Amsterdam – Boston – Heidelberg – London – New York – Oxford – Paris – San Diego
San Francisco – Singapore – Sydney – Tokyo
ELSEVIER B.V.
Sara Burgerhartstraat 25
P.O. Box 211, 1000 AE
Amsterdam, The Netherlands
ELSEVIER Inc.
525 B Street, Suite 1900
San Diego, CA 92101-4495
USA
ELSEVIER Ltd
The Boulevard, Langford Lane
Kidlington, Oxford OX5 1GB
UK
ELSEVIER Ltd
84 Theobalds Road
London WC1X 8RR
UK
q 2004 Elsevier B.V. All rights reserved.
This work is protected under copyright by Elsevier B.V., and the following terms and conditions apply to its use:
Photocopying
Single photocopies of single chapters may be made for personal use as allowed by national copyright laws. Permission
of the Publisher and payment of a fee is required for all other photocopying, including multiple or systematic copying,
copying for advertising or promotional purposes, resale, and all forms of document delivery. Special rates are
available for educational institutions that wish to make photocopies for non-profit educational classroom use.
Permissions may be sought directly from Elsevier’s Rights Department in Oxford, UK: phone (+44) 1865 843830, fax
(+44) 1865 853333, e-mail: [email protected]. Requests may also be completed on-line via the Elsevier
homepage (http://www.elsevier.com/locate/permissions).
In the USA, users may clear permissions and make payments through the Copyright Clearance Center, Inc., 222
Rosewood Drive, Danvers, MA 01923, USA; phone: (+1) (978) 7508400, fax: (+1) (978) 7504744, and in the UK
through the Copyright Licensing Agency Rapid Clearance Service (CLARCS), 90 Tottenham Court Road, London
W1P 0LP, UK; phone: (+44) 20 7631 5555; fax: (+44) 20 7631 5500. Other countries may have a local reprographic
rights agency for payments.
Derivative Works
Tables of contents may be reproduced for internal circulation, but permission of the Publisher is required for external
resale or distribution of such material. Permission of the Publisher is required for all other derivative works, including
compilations and translations.
Electronic Storage or Usage
Permission of the Publisher is required to store or use electronically any material contained in this work, including
any chapter or part of a chapter.
Except as outlined above, no part of this work may be reproduced, stored in a retrieval system or transmitted in any
form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written
permission of the Publisher.
Address permissions requests to: Elsevier’s Rights Department, at the fax and e-mail addresses noted above.
Notice
No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter of
products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or
ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent
verification of diagnoses and drug dosages should be made.
First edition 2004
Library of Congress Cataloging in Publication Data
A catalog record is available from the Library of Congress.
British Library Cataloguing in Publication Data
The political economy of trade, aid and foreign investment
policies
1. Commercial policy 2. Investments, Foreign – Government
policy 3. International economic relations
I. Mitra, Devashish II. Panagariya, Arvind
382.3
ISBN: 0-444-51597-6
W1 The paper used in this publication meets the requirements of ANSI/NISO Z39.48-1992 (Permanence of Paper).
Printed in The Netherlands.
Foreword
During the last two decades, the political economy of trade policy has emerged as
one of the most important branches of international trade. It is therefore a great
pleasure for me to see the leading scholars Devashish Mitra and Arvind
Panagariya bring together a team of relatively eminent scholars to shed light on a
number of important problems in this area. The volume contains a dozen essays
that push the frontiers of the political economy literature on trade, aid and
investment policy in a major way. It sheds new light on questions such as how
trade policy is formed in the presence of parliamentary and congressional
institutions, why trade policy formulation often sacrifices the terms of trade gains,
whether the politicians’ ideology or constituent interests have the determining role
in policy formulation, whether lobbying takes place along industry or factor lines,
how liberalization by one country can induce liberalization by its trading partners,
and how imperfect competition and quantitative restrictions interact to affect the
political feasibility of free trade areas. The collection offers an excellent balance
between theory and empirical analysis.
I am pleased the volume is dedicated to Ed Tower, who is a first-rate trade
theorist and has made important contributions to both the normative and positive
branches of trade theory and also the empirical trade literature. I have no doubt
scholars as well as trade policy analysts will greatly benefit from the essays.
Jagdish Bhagwati
University Professor,
Columbia University
Preface
In this volume, a group of distinguished scholars in international trade,
especially from the younger generation, analyzes several important but
neglected aspects of the political economy of international economic policy.
Apart from the Introduction, the volume contains 12 essays that make major
advances in the area of political economy of trade, aid and investment policy. In
selecting the essays, we have sought a balance between theory and empirical
work. And even though political economy analysis is "positive" by nature, many
of the articles carry "normative" implications and therefore lessons for policy
makers.
The volume is dedicated to our friend Ed Tower, Professor of Economics,
Duke University. While Ed’s many contributions to trade theory and his
encouragement to younger scholars are reason enough to dedicate a volume
that predominantly contains essays by younger scholars, there is also some
history behind it. In November 1999, one of us (Panagariya), who had
benefited immensely from the encouragement given by Ed during the early
years of his career, came to know that Ed had been diagnosed with terminal
cancer. There were no proven treatments for the disease and Ed had decided to
go for an experimental treatment at the M.D. Anderson Cancer Center in
Houston, Texas.
Soon after the treatment began, Ed himself described the prognosis in an email
sent to David Feldman, his student and a contributor to this volume, as follows:
"The odds seem to be 38% probability that I can avoid a bone marrow transplant
and still go into remission and stay there for three years. They are 80% if one
includes the possibility of a bone marrow transplant. That seemed lots better than
anyone else was offering. Susan [whom Ed had married only four months earlier
in August 1999] is bearing up well."
Thus, even the optimistic prognosis was not very encouraging. The idea for the
volume gelled at that time with the hope that we could quickly do something while
Ed lived. Ed himself actively participated in the discussions, making numerous
suggestions and even offering to contribute to the volume. That reflected Ed’s
eternal optimism that he will be there when we complete the volume even at our
retarded pace. And no doubt, it was this optimism that allowed him to beat all odds
and, with the help of an experimental drug that turned out to be tailor-made for
him, he recovered fully. So it is with double pleasure and hope that he will be
around to do a similar volume to honor us on our 60th birthdays that we dedicate
this one to Ed!
Devashish Mitra
Syracuse, NY, USA
January, 2004
Arvind Panagariya
New York, NY, USA
January, 2004
x
Contributors
Alok Bohara
Professor of Economics
University of New Mexico
Carl Davidson
Professor of Economics
Michigan State University
David Feldman
Professor of Economics
College of William and Mary
Kishore Gawande
Roy and Helen Ryu Professor of Economics & Government
Texas A&M University
Amy Glass
Assistant Professor of Economics
Texas A&M University
Omer Gokcekus
Associate Professor, Whitehead School of Diplomacy & International Relations
Seton Hall University
Keith Hall
Chief Economist
US Department of Commerce
William Kaempfer
Professor of Economics and Associate Vice-Chancellor
University of Colorado - Boulder
Bilgehan Karabay
Economics PhD Student
University of Virginia
Justin Knowles
Graduate Student, Wharton School
University of Pennsylvania
Pravin Krishna
Professor of Economics, Brown University and
Faculty Research Fellow, The National Bureau of Economic Research
Philip I. Levy
Senior Staff Economist
Council of Economic Advisers, The White House
Christopher S. Magee
Assistant Professor of Economics
Bucknell University
Steven J. Matusz
Professor of Economics
Michigan State University
Wolfgang Mayer
David Sinton Professor of Economics
University of Cincinnati
John McLaren
Professor of Economics, University of Virginia and
Faculty Research Fellow, The National Bureau of Economic Research
Devashish Mitra
Associate Professor of Economics, Syracuse University and
Faculty Research Fellow, The National Bureau of Economic Research
Alex Mourmouras
Senior Economist
International Monetary Fund
Douglas Nelson
Professor of Economics, Tulane University and
Professorial Research Fellow, Leverhulme Centre for Research
on Globalisation and Economic Policy,
University of Nottingham
Arvind Panagariya
Jagdish Bhagwati Professor of Indian Political Economy and
Professor of Economics
Columbia University
xii
Martin Richardson
Professor of Economics
Australian National University
Martin Ross
Economist
Research Triangle Institute
Kamal Saggi
Professor of Economics
Southern Methodist University
Ed Tower
Professor of Economics
Duke University
xiii
Contents
Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
Chapter 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Devashish Mitra and Arvind Panagariya
1.1 Part I. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Part II. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Part III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Part IV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Part I. Making of Trade Policy: Theory
Chapter 2 Trade Policy Making by an Assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
John McLaren and Bilgehan Karabay
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 A specific-factors model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.1 Basic structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.2 Voters’ preferences and equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.3 Comparative statistics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3 A Mayer–Heckscher–Ohlin Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4 The electoral college . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.5 Conclusions and open questions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Chapter 3 Should Policy Makers be Concerned About Adjustment Costs? . . 31
Carl Davidson and Steven J. Matusz
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 The model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.2.2 Formalizing the model and finding the initial steady state . . . . . . . 37
3.2.3 Adjustment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.2.4 Strengths and weaknesses of our model. . . . . . . . . . . . . . . . . . . . . . . . . 47
3.3 Aggregate adjustment costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.4 Adjustment costs and labor market flexibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Appendix B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Appendix C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Chapter 4 Non-Tariff Barriers as a Test of Political Economy Theories . . . . 69
Philip I. Levy
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.2 Developing testable implications of the theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.3 Empirical evidence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.3.1 Existing tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.3.2 Instrument choice as a test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.3.3 Evidence on non-tariff barriers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.4 Alternative explanations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Chapter 5 The Peculiar Political Economy of NAFTA:
Complexity, Uncertainty and Footloose Policy Preferences . . . . . . 91
H. Keith Hall and Douglas R. Nelson
5.1 The fact: footloose aggregate preferences on NAFTA. . . . . . . . . . . . . . . . . . . . 93
5.2 Policy complexity, social learning and footloose preferences. . . . . . . . . . . . . . 99
5.3 An illustrative model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.4 Conclusion: on economists as participants in the politics
of trade policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
Part II. Making of Trade Policy: Empirical Analysis
Chapter 6 Interest and Ideology in the 1988 Omnibus Trade Act:
A Bayesian Multivariate Probit Analysis . . . . . . . . . . . . . . . . . . . . . . . 113
Alok K. Bohara and Kishore Gawande
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.2 Background: the Trade Omnibus Act. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
xvi
6.3 Analysis of the multivariate probit model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.4 Theory and measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.5 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
6.6 Empirical analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
6.6.1 Substantive issues of focus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
6.6.2 Analysis of MVP estimates from HOUSE voting . . . . . . . . . . . . . . 131
6.6.3 Analysis of MVP estimates from SENATE voting . . . . . . . . . . . . . 137
6.6.4 Sensitivity analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.7 Model comparisons: Ideology versus interest . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
6.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Appendix A: Bayesian analysis of the multivariate probit model. . . . . . . . . . . . . . . . 151
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Chapter 7 Industry and Factor Linkages Between Lobby Groups . . . . . . . . . 159
Christopher Magee
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
7.2 Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
7.3 Empirical analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
7.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Chapter 8 Sweetening the Pot: How American Sugar Buys Protection . . . . 177
Omer Gokcekus, Justin Knowles and Edward Tower
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
8.2 The US sugar program. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
8.2.1 Price support loans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
8.2.2 Import restrictions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
8.3 Consequences of the program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
8.3.1 Increases users’ costs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
8.3.2 Benefits for producers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
8.3.3 Net effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
8.4 Sugar interest groups’ contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
8.5 Determinants of sugar’s campaign contributions . . . . . . . . . . . . . . . . . . . . . . . . 184
8.6 Marginal effects of different attributes on the probability of getting
money and the amount of money received by incumbent senators . . . . . . . 188
8.7 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
xvii
Part III. Inter-Country Interactions
Chapter 9 Unilateralism in Trade Policy: A Survey of Alternative
Political-Economy Approaches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Pravin Krishna and Devashish Mitra
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
9.2 Models of reciprocated unilateralism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
9.2.1 A median-voter model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
9.2.2 A model with trade policy lobbying and endogenous
lobby formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
9.2.3 A model of leadership in trade policy negotiations . . . . . . . . . . . . . 208
9.3 Models of endogenous unilateralism. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
9.3.1 Unilateral commitment to free trade as a means of
preventing capital misallocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
9.3.2 Unilateral commitment to free trade as a means of
preventing wasteful political (organizational) activity . . . . . . . . . . 209
9.4 Concluding remarks: the current state of the literature and issues
for future research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
Chapter 10 Trade Creation and Residual Quota Protection in a
Free Trade Area with Domestic Monopoly . . . . . . . . . . . . . . . . . . . . . 213
David H. Feldman and Martin Richardson
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
10.2 Single domestic firm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
10.2.1 An FTA may not lead to any trade creation . . . . . . . . . . . . . . . . . . . 217
10.2.2 Welfare effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
10.3 The political economy of FTA formation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
10.3.1 FTA is politically unattractive when profits and revenues
are equally valued . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
10.3.2 FTA can only be politically attractive if profits are more
highly valued than revenues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
10.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Chapter 11 The Political Economy of Trade Sanctions Against
South Africa: A Gravity Model Approach . . . . . . . . . . . . . . . . . . . . . . 233
William H. Kaempfer and Martin Ross
11.1 Introduction: the law of demand for sanctions . . . . . . . . . . . . . . . . . . . . . . . . . . 234
11.2 South African trade during the 1980S . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
xviii
11.3 The gravity model of trade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
11.4 Empirical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
11.5 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
Part IV. Aid and Foreign Investment Policies
Chapter 12 The Political Economy of Unconditional and Conditional
Foreign Assistance: Grants vs. Loan Rollovers . . . . . . . . . . . . . . . . . 249
Wolfgang Mayer and Alex Mourmouras
12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
12.2 The common agency model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
12.2.1 Decision makers and their objectives . . . . . . . . . . . . . . . . . . . . . . . . . . 254
12.2.2 Political equilibrium with unconditional assistance. . . . . . . . . . . . . 256
12.2.3 Political equilibrium with conditional assistance . . . . . . . . . . . . . . . 257
12.3 Instruments of assistance: loan rollovers vs. a final grant . . . . . . . . . . . . . . . . 260
12.4 Unconditional assistance: the IFI should use a grant . . . . . . . . . . . . . . . . . . . . 262
12.4.1 Unconditional loan decisions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
12.4.2 Unconditional grant decisions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
12.5 Conditional assistance: the IFI should use loans . . . . . . . . . . . . . . . . . . . . . . . . 266
12.5.1 Conditional loan decisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
12.5.2 Conditional grant decisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
12.6 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
Appendix A. Derivation of Equation 18 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
Appendix B. Derivation of Equation 19 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
Chapter 13 Crowding Out and Distributional Effects of FDI Policies . . . . . . . 277
Amy Jocelyn Glass and Kamal Saggi
13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
13.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
13.3 No intervention and national treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
13.4 Discriminatory treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
13.4.1 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285
13.4.2 Policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
13.4.3 Discriminatory versus national treatment . . . . . . . . . . . . . . . . . . . . . . 288
13.5 Most-favored-nation treatment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
13.5.1 Equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
13.5.2 Policy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
xix
13.5.3 Discriminatory versus MFN treatment . . . . . . . . . . . . . . . . . . . . . . . . 290
13.5.4 MFN versus national treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
13.6 Another basis for discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
13.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
Appendix A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
A.1. Proof of Proposition 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
A.2. Proof of Proposition 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
A.3. Proof of Proposition 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
A.4. Proof of Proposition 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
A.5. Proof of Proposition 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
A.6. Proof of Proposition 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
xx
CHAPTER 1
Introduction
DEVASHISH MITRAa,b,* and ARVIND PANAGARIYAc
aDepartment of Economics, The Maxwell School of Citizenship and Public Affairs,
Syracuse University, Eggers Hall, Syracuse, NY 13244, USAbNational Bureau of Economic Research, 1050 Massachusetts Avenue, Cambridge,
MA 02138-5398, USAcSchool of International and Public Affairs, International Affairs Building, Columbia University,
New York, NY 10027, USA
Economic policies, in practice, often deviate from what economists regard as
optimal. The principal reason behind this persistent empirical regularity is that
distributional concerns dominate the policy-formulation process through both
general-interest and special-interest politics. While the former works through the
government’s attempts to obtain the support of the majority in an inherently
unequal society, the latter works through the “sale” of policies to powerful interest
groups called “lobbies” in the political-economy literature. The theoretical and
empirical analysis of these channels of influence on economic policy is the branch
of economics that has been labeled “Political Economy”. This branch focuses
primarily on the “positive” rather than “normative” aspects of policy. In this
volume, we focus on the political economy of international economic policy, with
special emphasis on trade policy. While the proposition that trade barriers by a
country are harmful to its own overall well being is not controversial among
policy analysts, national governments throughout history have rarely embraced
*Corresponding author. Address: Department of Economics, The Maxwell School of Citizenship and
Public Affairs, Syracuse University, Eggers Hall, Syracuse, NY 13244, USA.
E-mail address: [email protected]
free trade. The explanation of why this is so has been the preoccupation of the
political-economy theories that have assumed the center stage in the field of
international theory in the last two decades. These theories broadly consist of
majority-voting (general-interest politics) and lobbying (special-interest politics)
models.
While considerable progress has been made in formalizing the process of
international economic policy formulation, the existing literature remains
deficient in several respects. First, the existing models take a relatively simplistic
view of the political-economy environment facing various agents. For example, in
contrast to the observed reality, a large majority of the models view the
government as a monolithic entity. Second, there is only limited recognition of the
political-economy interactions between interest groups across national borders in
the existing literature. Third, dynamic factors governing trade policy formulation
have been essentially absent. Fourth, learning and imperfect information
regarding the costs and benefits of trade policy are almost never incorporated
into political economy models. Fifth, the empirical work in this area, both cross-
industry within national boundaries and cross-national, is very much in infancy
and in need of further development. Sixth, other aspects of international economic
policy such as foreign aid and foreign direct investment (FDI) have been scarcely
addressed in the existing literature. This is a serious lacuna since the trade reform
in developing countries is frequently tied to aid from international financial
institutions (IFIs) such as the World Bank and International Monetary Fund.
Likewise, reforms of trade and direct foreign investment policies go hand in hand.
In this volume, a group of distinguished scholars in international trade analyzes
these and related issues. All essays are original. The volume contains 12 essays
topped by an introduction by the editors.
As explained above and described below in detail, the volume makes major
advances in the area of political economy of trade, aid and foreign investment
policies. We, therefore, expect it to be of considerable interest to academic
researchers and students of international economics. Because of this volume’s
obvious focus on the process of policy formulation, we believe that it will also
interest economists at think tanks, international institutions such as the World
Bank, the International Monetary Fund and the World Trade Organization, and
trade policy analysts in the developed and developing countries. Finally, we
expect many of the articles in this volume to be an integral part of graduate
reading lists in general courses in international trade and more specialized
political economy courses.
Setting aside the introduction, the volume is divided into four fairly distinct
parts. The essays in Part I look at the formation of trade policy and its effects on
different sections of the society within a single economy context. The essays in
D. Mitra and A. Panagariya2
Part II consider the same issue from empirical perspectives. Part III focuses on
inter-country interactions in the presence of endogenous trade policy. Finally, Part
IV considers issues relating to foreign aid and investment policies. In the
following, we offer a brief description of the 12 essays in the volume.
1.1. PART I
In the existing political economy literature on trade policy determination, a single,
monolithic policy maker is often assumed. In Chapter 2, McLaren and Karabay
make a departure from such a simple structure to study trade policy setting in the
presence of parliamentary or congressional institutions. They also incorporate
electoral competition between political parties and show that in their setting the
relationship between the likelihood of import protection and the geographical
concentration of import-competing interests is non-monotonic with a maximum
occurring at moderate levels of concentration. Too much concentration leads to a
control of too few seats, while too much dispersion leads to no control of any
seats. Thus, they derive an empirically testable relationship and therefore, in their
chapter discuss the implications of this model for empirical work. They also
discuss the applicability of their analysis to the American polity.
While Chapter 2 departs from the monolithic policymaker setting, Chapter 3
makes an innovation in another direction. It deals with dynamic considerations
and adjustment costs in the determination of short-run losses versus long-run gains
from trade reforms, which should in turn affect the support for such reforms.
Davidson and Matusz construct a dynamic, general equilibrium trade model with
time and resource costs for retraining. Interestingly, some of the key parameters of
their model such as labor turnover rates are observable in the real world. Based on
true estimates of such parameters and using various alternative estimates for their
other less observable parameters, they calibrate their model. The authors show
that short-run losses could, in the aggregate, be quite large and can be significant
relative to the long-run benefits from such reforms. Not surprisingly, economies
with more flexible labor markets are found to suffer lower adjustment costs and
obtain higher gains from liberalization. Finally, Davidson and Matusz emphasize
the role of the government in devising policies to compensate groups that bear the
bulk of the short-run adjustment costs, thereby making trade reforms politically
more feasible.
From dynamic considerations, we move on to Chapter 4 by Phil Levy that offers
a compelling critique of the current political economy models of trade policy.
Levy’s evaluation of the current literature is based on a look at the evidence on the
nature of trade policy instruments used in practice. The author focuses on political
Introduction 3
economy models in which governments actually or effectively maximize
weighted social welfare functions. This includes the state-of-the-art model of
Grossman and Helpman (1994).1 Levy argues that whether or not the government
is the single domestic player or there are other players (such as lobbies) involved
the government ultimately acts as a “unitary player” in international interactions
in such models. Besides, recent theoretical research has demonstrated that such
unitary actors care exclusively about terms of trade in international negotiations.
Levy then goes on to argue that the structure of United States protection, biased
towards the use of non-tariff barriers that sacrifice terms of trade, calls into serious
question this monolithic view of the government. He, then, goes on to discuss, in
an informal way various possible, alternative theories of political economy that
could accommodate this stylized fact.
In Chapter 5, Hall and Nelson examine another interesting aspect of the
political economy of US trade policy, namely the “apparent footlooseness of
aggregate opinion” on the North American Free Trade Agreement (NAFTA).
More specifically, they first document the massive shift in US public opinion on
the NAFTA between the early and the mid-1990s in the absence of any change in
economic and political changes in the US during that period. They then present an
illustrative model of footloose policy preferences. Within a Heckscher–Ohlin
model, they analyze the changing support for the formation of an free trade
agreement (FTA) in the presence of imperfect public and private information on
the characteristics of the other member countries of the potential FTA. Actions of
other citizens in the home country in terms of support for or against the FTA can
release information about the characteristics of the partner FTA countries and
about the potential benefits from the FTA. Small amounts of information released
can start or reverse information cascades in support for or against a FTA. Thus,
this model of political economy provides a greater role for economic policy
analysts in the determination of policy than do most standard models since it
allows for additions by expert economists to the flow of information subject to
social learning.
1.2. PART II
Part II tackles the political economy of trade policy empirically. Through rigorous
econometric analysis, the authors look for the deep determinants of trade policy in
the real world. They also look for robust empirical regularities in protection as
well as in lobbying activity. In Chapter 6, Bohara and Gawande focus on
1 See also Grossman and Helpman (1995).
D. Mitra and A. Panagariya4
congressional voting on which there is difference of opinion among scholars as to
whether its main determinant is the ideology of the politician or the constituents’
interests or a combination of the two. This chapter undertakes (a) conventional
testing of models of interest and ideology from size and signs of coefficients; (b)
comparison of non-nested models of interest and ideology for House and Senate
separately; and (c) examination of inter-chamber heterogeneity in voting
behavior. The authors estimate multivariate probit models, three each for the
House and the Senate, using Gibbs sampling together with the Metropolis–
Hasting algorithm for assessing the determinants of sets of key roll call votes
during the legislation of the Omnibus Trade and Competitiveness Act, 1987–
1988. Based on conventional testing, all models of interest and ideology are
affirmed for at least some of the votes analyzed, and some models for all the votes
analyzed. Based on model comparisons, interest dominates ideology in Senate
voting in all three models, while ideology dominates House voting in two of three
models, holding party neutral. If party is taken to represent interest, then interest
dominates ideology in both chambers.
The next chapter (Chapter 7) focuses on the actual lobbying activity as
measured by political contributions. Magee empirically investigates whether
lobbying takes place along factor or industry lines. The Heckscher–Ohlin model,
based on perfect inter-sectoral factor mobility, predicts that across all industries
lobby groups representing a particular factor have common interests with respect
to decisions on campaign contributions. Under the specific factors model, it is the
industry of origin that matters. Econometric analysis on the data on campaign
contributions during the 1991–1992 election cycle in the US provides evidence
for imperfect but some factor mobility. Magee finds significant, positive
correlations in both cross-industry, within factor and cross-factor, within industry
spending patterns of political action committees.
Chapter 8 is an in-depth case study of the protection in the US sugar industry,
one of the most politically active industries in the country. Gokcekus, Knowles
and Tower present an empirical study of the capture of rents by sugar growers
from the US sugar program. Surprisingly, politicians are almost always against
phasing this program out in spite of the huge excess burden it creates on society.
Using Tobit analysis, the authors find that the power and willingness of politicians
to protect are key determinants of the campaign contributions they receive, with
significant premiums on membership on the Senate Agriculture, Nutrition and
Forestry Committee and the Agricultural Production, Marketing, and Stabilization
of Prices Subcommittee. Being a member of the party in power also can draw
additional campaign money. Finally, whether the politician belongs to a sugar or
non-sugar state also matters.
Introduction 5
1.3. PART III
In Part III, authors go beyond domestic politics and explicitly introduce
international political interactions. In this context, Chapter 9 by Krishna and
Mitra on unilateralism surveys their earlier work on different channels through
which trade policy in one country will have an impact on trade policy
determination in its partner country and conditions under which there will be
strategic complements meaning liberalization by one will induce liberalization by
the other, as argued very eloquently and clearly by Jagdish Bhagwati in many of
his writings.2 The first channel the authors analyze is through lobby formation.
They argue that trade liberalization in one country, by expanding the potential
market for the output of exporters in its partner country, increases the benefits for
these exporters from free trade. This in turn means that these exporters now have a
greater incentive to form a lobby. This new lobby is then able to neutralize the
import-competing lobby, leading to a trade reform in this country as well. The
next argument relies on majority voting. Trade liberalization in one country
expands its partner country’s export sector relative to the import-competing
sector, through a movement of producers/factors from the latter to the former.
Since producers in the export sector benefit from trade reforms in their own
country, there is now greater support for such reforms. Thus, trade reforms in one
country will make reforms more likely in its partner country. This also means that
there is the possibility of multiple equilibria—protectionism in both countries and
free trade in both countries. The authors also review work by Coates and Ludema
(2001) on leadership in trade policy negotiations as well as models of endogenous
unilateralism as, in for example, Maggi and Rodriguez-Clare (1998).
Chapter 10 by Feldman and Richardson explores the consequences of
discriminatory trade liberalization in the presence of quota-protected single-firm
industries. In the earlier literature we see that a quota-protected small country with
a competitive market gains from an FTA with a partner that supplies the good at a
lower price. The authors here show that with a domestic industry composed of a
single firm, an FTA must be welfare improving without necessarily leading to
trade creation. Under these conditions, the authors study the political support for
and against the formation of an FTA. They also analyze the incentives to liberalize
the initial quota upon formation of an FTA.
In Chapter 11, Kaempfer and Ross study another aspect of inter-country
interactions, namely, the political economy of trade sanctions. They use the
gravity model for their analysis. They hypothesize that the severity of the
application of anti-apartheid trade sanctions, that led to the dramatic fall of
2 See, for instance, Bhagwati (2002).
D. Mitra and A. Panagariya6
South African trade during the mid-1980s, was inversely related to the cost of
imposing those sanctions—a law of demand for sanctions. Therefore, this cost was
positively related to negative determinants of trade (the determinants of trading
costs such as distance) and negatively related to its positive determinants (such as
country size in the gravity model). The results from the authors’ estimation of a
gravity model of trade flows as a function of distance and economic mass for
South Africa show an improved explanatory power for the model after the world-
wide application of sanctions in 1986. Most importantly, the influence of distance
on trade flows increases significantly after the sanctions are enacted. These
findings are consistent with their law of demand for sanctions, thereby suggesting
that the restrictiveness of sanctions was inversely related to the importance of a
country’s trade with South Africa, driven by its relative size or distance (from
South Africa).
1.4. PART IV
The last part of the book focuses on two very important but highly neglected
aspects of international economic policy in the political economy literature,
namely foreign aid and foreign investment. In an innovative, promising line of
research in Chapter 12, Mayer and Mourmouras focus on conditional foreign aid
granted by IFIs to developing countries. They allow for three types of actors in
their multi-period political-economy model: the government, a domestic interest
group, and an IFI. The domestic government’s objective is to maximize the
present value of a weighted sum of political contributions and national welfare.
The IFI is assumed to have only indirect influence on the government, through the
impact of its foreign aid package on national welfare. The domestic interest group
benefits from distortionary economic policies on which it conditions its political
contributions. The IFI’s objective function is assumed to be the present value of a
weighted sum of the national welfare levels of both the assistance-receiving
country and the assistance-financing rest of the world. Under these assumptions,
the authors shed light on criteria for choosing the form of IFI assistance—grants or
loans. A one-time grant in the initial period is a more efficient instrument than
loans when assistance is not conditioned on the adoption of distortion-reducing
policies. Rollover loans, on the other hand, are more efficient than a grant when
assistance is conditional. The reason for the reversal in ranking is that
conditionality enables the IFI to achieve Pareto optimality, and while conditions
on policies can be enforced with renewable loans, they cannot be enforced over
time with a one-time grant.
Introduction 7
The last chapter in the volume by Glass and Saggi focuses on the political
economy of foreign investment policies. The authors examine the effects of
international investment agreements and the rationale for the “most favored
nation” and “national treatments” clauses in such agreements. They find that
increasing the tax on firms from a particular source country leads to substitution of
FDI from that country by those from other countries. This also pushes the wages in
the more highly taxed country down relative to other source countries. When the
host country has full freedom to discriminate, it exercises it by levying a larger tax
on multinationals that have more to gain from FDI. Such multinationals therefore
would gain, at the cost of those that have a smaller desire to engage in FDI, from
both “most-favored nation” and “national treatments” standards, requiring non-
discrimination relative to other foreign firms and to domestic firms, respectively.
Thus, such clauses might increase the aggregate amount of FDI. The Glass–Saggi
analysis, therefore, can clearly identify the winners and losers from these clauses
and enable us to throw some light on the emergence of plausible interest groups
for and against investment agreements incorporating them.
The volume, thus, offers an assortment of contributions on the political
economy of various aspects of international economic policy. We also believe
these contributions have made significant advances over the existing literature,
extending it in several new directions and incorporating many ignored but
important and realistic elements into it.
REFERENCES
Bhagwati, J. (2002). Free Trade Today, Princeton, NJ: Princeton University Press.
Coates, D. and Ludema, R. (2001). A theory of trade policy leadership. Journal of
Development Economics, 65(1), 1–29.
Grossman, G. and Helpman, E. (1994). Protection for sale. American Economic Review, 84,
833–850.
Grossman, G. and Helpman, E. (1995). Trade wars and trade talks. Journal of Political
Economy, 103, 675–708.
Maggi, G. and Rodriguez-Clare, A. (1998). The value of trade agreements in the presence
of political pressures. Journal of Political Economy, 106(3), 574–601.
D. Mitra and A. Panagariya8
CHAPTER 2
Trade Policy Making by an Assembly
JOHN McLAREN,* BILGEHAN KARABAY,
Department of Economics, Rouss Hall, University of Virginia,
Charlottesville, VA 22903-3288, USA
Abstract
Economists’ models of trade-policy determination generally assume unitary
government. We offer a congressional model. Under assumptions guaranteeing a
median-voter outcome under a unitary model, we find a wide range of possible
outcomes: any policy from the 25th to the 75th percentile voter’s optimum can
emerge in equilibrium, depending on how voters are divided up into voting
districts. The equilibrium policy is the optimum of the median voter of the median
district. Protection is most likely if import-competing interests are not too
geographically concentrated or too disperse. We discuss implications for the
American electoral college system, and for empirical work.
Keywords: Congress, international trade policy, political economy
JEL classifications: D72, P16
2.1. INTRODUCTION
Increasingly, trade economists have shown an interest in understanding the
determinants of trade policies as well as their effects. Influential examples
include Mayer’s (1984) electoral model of trade policy formation, Findlay and
*Corresponding author.
E-mail address: [email protected]
Wellisz’s (1982) model of tariff lobbying and Grossman and Helpman’s
(1994) influence-peddling model of tariff setting. The literature has grown
quite dense in recent years; see Nelson (2002) for an interpretative survey.
Despite the research interest, the theoretical literature has remained strikingly
unidimensional in an important respect: the assumption of a unitary government.
If Mayer’s model is interpreted as a contest between candidates for office (who
commit to policy decisions in advance of elections), then once the winning
candidate takes office he/she sets the tariff without any need for consultation. In
lobbying and influence-peddling models, an interest group influences a decision
maker who is assumed to have unimpeded power to set trade policy within the
country. These assumptions have become standard practice.
The unreality of these assumptions is revealed by a glance at trade policy
history. With the possible exception of pure administered protection such as anti-
dumping, trade policy in democracies is normally the product of multiple decision
makers, often with sharply differing interests. In most countries trade policy is set
by a parliament and is therefore the outcome of legislative bargaining and
cooperative or non-cooperative voting. In the United States, it is set by two houses
that must come to mutual agreement, and is then subject to presidential veto. In all
democracies, a trade treaty is negotiated by the executive branch and must then be
ratified by a domestic assembly.
All this requires that we think of trade policy as being set by an organization,
not by a single individual endowed with authority. Many details of trade policy
making in practice cannot even be addressed without such considerations, such as
the battles for ratification of the NAFTA and Uruguay round in the United States
and the central issue of “fast-track” authority, which is meaningless in a pure
presidential model.
In this chapter, we extend one standard model of trade policy formation
(specifically the median-voter framework most associated with Mayer (1984)) to a
rudimentarymodel of a government by assembly inwhich political parties compete
for control of seats by making binding election promises. We focus on the simplest
possible example in the hope that it will make some of the key issues as clear as
possible. This exercise reveals a number of sharp predictions. First, import-
competing interests are more likely to receive protection if they are moderately
geographically concentrated. If import-competing interests are concentrated in a
few locations in the country, they may dominate those areas politically but will
control too few seats to be able to control the assembly. If they are too disperse, they
will not be politically dominant anywhere and will thus control no seats. Only with
a moderate level of geographical concentration can they secure enough political
clout for protection. This non-monotonic relationship should be readily testable.
J. McLaren and B. Karabay12
Second, an assembly system will be less likely to secure protection than a
presidential system if import-competing interests are in the majority
nationally, and more likely if they are a national minority. If we interpret
the second case as more likely in practice, this argues for a presumption that
assemblies tend to be more protectionist than presidential systems.
Third, the unique equilibrium tariff is the optimal tariff of the median voter in
the median congressional district, rather than the national median voter. This
results in a dramatic break with the familiar models: for a given national
distribution of trade policy preferences, depending on the way voters are
allocated to voting districts, the equilibrium tariff can be anywhere from the 25th
percentile voter’s most preferred level to the 75th percentile voter’s most
preferred. Thus, moving from a single district (the unitary model) to two districts
changes the range of outcomes dramatically, while a subsequent increase in the
number of districts does not change the range at all. This indicates that the
median voter results are actually quite fragile.
This chapter is related to a number of strains of existing literature. Political
science has, of course, no habit of assuming a unitary government. There is a long
history of political scholarship on the behavior of Congress; influential studies
include Krehbiel (1991) and Poole and Rosenthal (1997). In recent years, many
political scholars have focused on intra-governmental complications in the
formation of trade policy. Putnam (1988), for example, studies “two-level games”,
in which one branch of government must negotiate with a foreign government and
then present the agreement to domestic agents for ratification. Trade treaties are
naturally a prime example. Lohmann and O’Halloran (1994) and Bailey et al.
(1997) look at congressional behavior in setting trade policy, focusing on the
relationship between executive and legislative branch behavior and the
interpretation of such institutions as the Reciprocal Trade Agreements Act
(RTAA) and the “fast track” authority, both of which were acts of congress that at
different times have constrained Congress’ ability to amend trade treaties.
A number of authors have looked closely at the behavior of congressional voting
on trade policy. For example, Baily and Brady (1998) and Dennis et al. (2000) both
study the importance of constituency characteristics including voter heterogeneity
for explaining how senators voted on various recent trade bills. Economists
studying congressional voting behavior include Peltzman (1985) (who showed that
economic interest variables matter muchmore in explaining votes on taxation once
state fixed effects are controlled for), Irwin and Krozner (1999) (who study the
postwar changes in Republican congressional voting on trade), and Baldwin and
Magee (2000) (who study congressional log-rolling on trade policy).
Most of this work focuses on fine details of political institutions such as
the RTAA or fast-track authority, or analyzes empirically how individual senators
Trade Policy Making by an Assembly 13
or representatives choose to vote. The present chapter, in the spirit of the
theoretical papers listed at the outset, begins with a very simple, abstract model, to
ask the question: how do economic fundamentals affect trade policy outcomes?
And how does that mapping change if we move from a unitary government model
to a government by assembly?
Section 2.2 describes the easiest form of assembly model, the “specific factors”
model in which each worker is qualified to work in only one sector. Section 2.3
presents a version with Heckscher–Ohlin features, which is in fact a
generalization of the main model in Mayer (1984). Section 2.4 shows how the
model can be adapted to analyze the effect of the “electoral college” system in the
United States. Section 2.5 offers some questions for future research.
2.2. A SPECIFIC-FACTORS MODEL
2.2.1. Basic structure
Consider an economy called Home with two sectors, X and Y ; each producing a
homogeneous good under competitive conditions. Good Y is the numeraire. The
only factor of production is labor, and each worker is either a “type X”, who can
produce X only, or a “type Y”, who can produce Y only. There is a continuum of
workers of type X; with measure LX ; and a continuum of type Y workers with
measure LY : These supplies are exogenously given and LY þ LX ¼ L: A worker of
type j can produce one unit of good j per hour.
All Home citizens have identical and homothetic preferences, with indirect
utility given by vðI; pÞ ¼ I=fðpÞ; where I denotes income, p denotes the price of
good X; and f is a price index, an increasing function of p:The world relative price of good X is denoted pW and is exogenous, since Home
is a small open economy. Assume that f0ðpWÞ=½f2 f0ðpWÞpW� . LX=LY : Theleft-hand side of this expression is (by Shephard’s lemma) the ratio of Home
demand for X to Home demand for Y at world prices, and the right-hand side is the
corresponding ratio of supplies. Thus, this condition ensures that Home has a
comparative advantage in Y:Every worker is a voter, and every voter is a worker. There are n districts, each
with the same number of voters. Each district will send a representative to an
assembly, which we will call the “congress”, and which will determine trade
policy. There are two parties, and all candidates for congress must be member of
one of these parties. Majority rule applies: the candidate with the largest number
of votes wins the seat (with coin flips to break a tie). Further, the party with the
larger number of seats can propose a trade policy; it goes up for a vote; and
J. McLaren and B. Karabay14
if it collects a majority of votes, it becomes law. Otherwise, the default of free
trade remains in effect. We assume that party leadership can impose loyalty on its
members, so that the majority party in congress effectively determines trade
policy. (If there is an exact tie in congress, a coin toss determines which party can
propose the trade policy, and it then goes up for a vote as before.)
An election is held to determine the representatives to congress. In each district,
each party fields exactly one candidate. The national leadership of each party
announces before the election what policy it will enact if it attains a majority in
congress. These announcements are made simultaneously. In each district, then,
each voter votes for the representative of the party whose announced policy that
voter prefers. (All voters vote, and there is no strategic voting; voters simply vote
their policy preferences, flipping a coin in the event of a tie.) Each party has an
objective function that is simply increasing in its expected number of seats in
congress.
2.2.2. Voters’ preferences and equilibrium
The X- and Y-voters are distributed to the various districts in an exogenous
pattern. Denote the fraction of voters in district i who are of type X by ri (sothat
P
i ri ¼ nLX=L and
P
i ð12 riÞ ¼ nLY=LÞ: Suppose that the only trade
policy instrument available is a tariff on good X; denoted in ad valorem terms
by t; and that the tariff cannot be negative (say, because an import subsidy
would create incentives for export and immediate re-import, which would be
difficult to police). Suppose further that the tariff revenues are distributed lump
sum to all workers equally. We can now show that the preferred policy of the Y
workers is free trade, while the preferred policy of the X workers is a strictly
positive tariff.
Letting M denote aggregate imports of X; the utility of an X-worker is
viðIx; pÞ ¼ Ix=fðpÞ;
where Ix ¼ pþ tpWM=L; the marginal value product of X labor plus the
typical X-worker’s share of tariff revenue. Recalling that p ¼ ð1þ tÞpW; we canwrite
›vðIX ; pÞ
›t¼
pW
wð pÞ1þ
M
Lþ
t›M
›tL
2 1þ tþtM
L
� �
pW w0ðpÞ
wðpÞ
2
664
3
775:
Trade Policy Making by an Assembly 15
If we evaluate this derivative at t ¼ 0; we get
›vðIX; pÞ
›t
����t¼0
¼pW
fðpWÞ1þ
M
L2 p
W w0ð pWÞ
wðpWÞ
" #
:
Note that by Shephard’s lemma pWf0ðpWÞ=fðpWÞ represents the share of the
X-worker’s expenditure that is spent on good X when t ¼ 0: It is therefore
between zero and unity, yielding
›vðIx; pÞ
›t
����t¼0
. 0:
Therefore, the X-worker’s most preferred tariff is strictly positive.
Depending on parameters, this most-preferred tariff could be prohibitive. Let �tdenote the prohibitive tariff, or the tariff rate such that M ¼ 0: Evaluating the
X-workers’ welfare derivative at that point
›vðIX ; pÞ
›t
����M¼0
¼pW
fðpÞ1þ
t›M
›tL
2 ð1þ tÞpWw0ðpÞ
wðpÞ
2
664
3
775
¼pW
fðpÞ1þ
t›M
›tL
2 aðpÞ
2
664
3
775;
where aðpÞ denote the share of consumer income spent on good X: (Strictlyspeaking, this needs to be interpreted as a left-hand derivative.) Since ›M=›t , 0;clearly, if good X has a sufficiently large budget share, the derivative will be
negative at the prohibitive tariff, and X workers will prefer a non-prohibitive tariff,
defined by ›vðIX ; pÞ=›t ¼ 0: Either way, denote the X-workers’ most preferred
tariff level by t ¼ tX:Treating the Y-workers in the same way, noting that the income of each
Y-worker is equal to Iy ; 1þ tpWM=L;we obtain:
›vðIY ; pÞ
›t¼
pW
fðpÞ
M
Lþ
t›M
›tL
2 1þ pW tM
L
� �w0ðpÞ
wðpÞ
2
664
3
775
¼pW
fðpÞ
M
Lþ
t›M
›tL
2 ðIY Þf0ðpÞ
fðpÞ
2
664
3
775:
J. McLaren and B. Karabay16
Note that IYf0ðpÞ=fðpÞ is each Y-worker’s consumption of good X; by Shephard’s
lemma. Further, using the same logic, M=L ¼ ð½LXIX þ LY IY �f0ðpÞ=fðpÞ2
LXÞ=L: Since X-workers cannot consume more of good X than they produce,
this cannot exceed LY IYf0ðpÞ=ðfðpÞLÞ: Since LY , L; we conclude that M=L ,
IYf0ðpÞ=fðpÞ: Since ›M=›t , 0; we conclude that Y-workers’ welfare is always
decreasing in the tariff, so their most preferred tariff is t Y ¼ 0:Before analyzing equilibrium in this model, let us consider what equilibrium
would be like in a more familiar, unitary government model. In other words, what
the equilibrium would be like if n ¼ 1: This is essentially the case of Mayer (1984)
(although the assumed structure of the economy is different), and it is well known
that the unique equilibrium in that case is that the median voter’s most preferred
tariff will be implemented. Thus, LX . LY implies t ¼ t X; and LX , LY implies
free trade.
However, the outcome is different if n . 1; as the following indicates.
Proposition 1. If there is an odd number of districts, then if they are ranked by
ri; if the median district has ri , 12; then the unique equilibrium in pure strategies
is one in which both parties commit to t ¼ 0; or free trade. If the median district
has ri . 12; then the equilibrium is t ¼ t x:
The proof is straightforward. Suppose that for the median district ri , 12: Then
if in equilibrium any party committed to a strictly positive tariff, the other party’s
best response would be a strictly lower tariff, ensuring a strict majority of votes in
a strict majority of districts. But then the first party’s tariff choice is sub-optimal,
since it could achieve an expected number of seats equal to n=2 by committing
itself to the same tariff as the other party. Thus, the only possible Nash equilibrium
involves both parties choosing a zero tariff. Further, since any deviation from the
zero tariff will only reduce the number of seats (to a certain minority rather than an
expected value of n=2); this is itself a Nash equilibrium. The proof for the case in
which ri . 12is parallel.
It is easy to deal with the cases in which the median-of-medians is not
unambiguously defined. Where n is odd, if ri ¼ 12; then any pair of tariffs in the
range ½0; tx� is an equilibrium. If n is even, and if i and iþ 1 are the middle two
districts in the ranking, then if ri; riþ1.
12; the equilibrium is t ¼ tx; if ri;
riþ1,
12; the equilibrium is free trade; and if ri # 1
2# riþ1; any pair of tariffs in
the range ½0; tx� is an equilibrium. These are all just generalizations of the median-
of-medians. We will henceforth ignore these knife-edge cases.
Although this is clearly a simple generalization of the median voter theory, it
should be pointed out that the difference in outcomes between the two models can
be large. Consider the following two polar cases. First, suppose that LX=L has
Trade Policy Making by an Assembly 17
a value slightly larger than 14; so that an X-worker would be nowhere near the
median voter and so under the unitary model we would clearly have free trade.
Now, in the congressional model suppose that n is fairly large and odd, and that the
X-workers are distributed evenly among ðnþ 1Þ=2 of the districts, with ri ¼ 0; inthe other districts. Now, a bare majority of the voters in those ðnþ 1Þ=2 districts
is of type X; and since this is the majority of the districts, the equilibrium is
now t ¼ tx:Second, suppose that LX=L has a value slightly smaller than 3
4; so that a Y-worker
would be nowhere near the median voter and under the unitary model we would
clearly have t ¼ tx: Now, in the congressional model suppose that the Y-workers
are distributed evenly among ðnþ 1Þ=2 of the districts, with ri ¼ 1 in the other
districts. Now, a bare majority of the voters in those ðnþ 1Þ=2 districts is of type Y;and since this is the majority of the districts, the equilibrium is now free trade.
In both these cases, the median voter is very different from the median–median
voter, and so the congressional model gives the opposite of the answer given by
the unitary model.
2.2.3. Comparative statistics
The following special case can help illustrate the role of intra-national geographic
distribution of industry on trade policy in this model. Suppose that there are two
kinds of district: there are m districts that have some X workers and some Y
workers, and there are n2 m districts that have only Y workers. For all of the
mixed districts, ri ¼ �r ; ðLX=LÞðn=mÞ: As m ranges from 1 to n; the distributionof import-competing workers becomes less concentrated and the number of
import-competing workers in each mixed district falls. To clear away some
taxonomy, assume that n is odd. Clearly, if m is less than ðnþ 1Þ=2; the outcome
will be free trade (because the median voter of the median district will be a Y
worker). At the same time, if ðLX=LÞ .12and m $ ðnþ 1Þ=2 the equilibrium will
be t ¼ tx; because even if the X-workers are spread as thinly as possible with
m ¼ n; the median voter in each district will be an X worker. If 2n=ððnþ 1ÞÞ �
ðLX=LÞ ,12; or ðLX=LÞ , ðnþ 1Þ=4n; the outcome will be free trade regardless of
m; since even if the X workers are as concentrated as possible subject to the
constraint that the median district be mixed (in other words, even if m ¼
ðnþ 1Þ=2Þ; the median voter in mixed districts will be a Y worker. Finally, if
ðnþ 1Þ=4n , ðLX=LÞ ,12; then if m , ðnþ 1Þ=2; the outcome will be free trade;
if m ¼ n; the outcome will be free trade; and for a range for m in between these
extremes the equilibrium will be t ¼ tx:To summarize, for this special case with homogeneous workers and
homogeneous capitalists, if import-competing workers are found in a majority
J. McLaren and B. Karabay18
of districts and the import-competing sector is neither so large that it dominates
the economy nor so small as to be politically negligible, then a protectionist policy
will emerge only if “m” is in a middle range, or only if the import-competing
workers are moderately geographically concentrated.
2.3. A MAYER–HECKSCHER–OHLIN MODEL
Now consider a different economy with the same political institutions. Here, we
study a version of the Mayer (1984) model, with the national assembly described
in Section 2.2 grafted onto it.
Consider an economy that produces two goods, X and Y ; using capital and laborwith constant-returns-to-scale technology. Both factors are homogeneous, and can
be transferred from production of one good to another instantly and costlessly.
There are therefore a single price w for labor and singe price r for capital services
throughout the economy. The aggregate amount of labor available is denoted by L;and the aggregate amount of capital is denoted by K: Each citizen i has K i units of
capital and one unit of labor. Capital endowments vary from person to person, and
the distribution can be summarized by a cumulative distribution function F: We
number the citizens in increasing order of wealth (so that a higher value of i
indicates a higher value of K i).
Good Y is the numeraire, pW is the world price of X; which is taken as given
because the economy in question is small, and p is the domestic price of X:Assume that X is labor intensive and that there are no factor-intensity reversals.
For now, we will assume for concreteness that X is the imported good (which is, of
course, the same as assuming that the country under consideration is capital
abundant compared to the rest of the world). The only trade policy available is
an ad valorem tariff t on imports of X; so that p ¼ ð1þ tÞpW: All tariff revenue isdistributed to the citizens in proportion to each citizen’s factor income. Thus, if
T is the aggregate tariff revenue, then citizen i receives a tariff revenue payment
of T i ¼ aiT ; where ai ¼ ðwþ rK iÞ=ðwLþ rKÞ: All citizens have identical and
homothetic preferences summarized by the indirect utility function vðp; IiÞ; whereIi represents the income of citizen i; including both factor income ðwþ rK iÞ and
redistributed tariff revenue T i:Mayer (1984) shows that in this framework, under weak conditions on utility
each citizen has strictly quasiconcave preferences over tariffs. We can thus speak
of the unique most-preferred tariff level ti for each citizen i:Mayer shows that for
a citizen i for whom K i ¼ K=L; ti ¼ 0: We can call this person the average
citizen, denoted �ı: Further, if K i. K�ı; ti , 0; while if K i
, K�ı; ti . 0:Essentially the poor want labor to be expensive, and thus want labor-intensive
Trade Policy Making by an Assembly 19
imports to be expensive, and hence desire trade protection. On the other hand, the
rich want capital to be expensive, and thus want labor-intensive imports to be
cheap, and hence desire subsidized imports if that is feasible. In addition, a citizen
with a higher value of K i will have a lower most preferred tariff, so by ranking the
citizens in increasing wealth, we rank them in decreasing order of desired
protection. (Of course, for a capital-poor economy, for which good X would be an
export good, these preferences and rankings would be reversed; the poor would
desire open trade, the rich protection, and ranking citizens by wealth would rank
them in increasing order of desired tariff.)
Now, to address the question of how tariffs are determined, we add to this model
the political structures as described in Section 2.2. Once again, we have n equally
sized districts, in each of which two candidates will compete for a seat in the
assembly. Each candidate belongs to one of the two national parties, which can
enforce national party discipline and can commit themselves in the election
campaign to future policy, where the majority party establishes the agenda and
determines the national tariff. The model studied by Mayer is essentially the
special case in which n ¼ 1; and the unique equilibrium is the implementation of
the most preferred tariff of the median voter (i.e., the voter i such that FðK iÞ ¼ 12Þ:
All workers are voters (whether or not they own capital), so the total number of
voters equals L:Some additional notation is necessary to characterize equilibrium in the case
with n . 2: Let the cumulative function for the K i values for the citizens in district
j be denoted by Fj: (Thus, of course, ð1=nÞP
j Fjð yÞ ¼ Fð yÞ ;y:Þ Let the most
preferred tariff of the national median voter be denote by tmed; and the most
preferred tariff of the median voter in district j be denoted by tmed; j: Without loss
of generality, let us number the districts in increasing order of tmed; j; and if n is
odd so that the median district is well-defined (i.e., district ðnþ 1Þ=2), then label
the most preferred tariff of the median voter of the median district tmed;med:It is straightforward, following the logic of the proposition of Section 2.2, to see
the following.
Proposition 2. If n is odd, then the unique equilibrium is for the most preferred
tariff of the median voter of the median district to be announced by both parties
and to be implemented. If n is even, then any pair of tariffs in the range ½tmed;n=2;tmed;n=2þ1� is an equilibrium.
The median-of-medians formula with n . 1 is clearly different from the simple
median of the n ¼ 1 case, but we need to be able to identify how different it is, and
especially how its empirically measurable properties differ. The first piece
of information to provide is a bound on how far from the simple median
J. McLaren and B. Karabay20
the median-of-medians can be. Consider the case in which n is odd. Note that one
half of the ðL=nÞ voters in the median district ðnþ 1Þ=2 have a higher K i; andhence desire a lower tariff, than the median voter of that district. In addition,
since tmed;ðn21Þ=2# tmed;ðnþ1Þ=2 ¼ tmed;med; at least one half of voters in district
ðn2 1Þ=2 prefer a lower tariff than the median-of medians. Following this logic
all the way down to district 1, we find that at least ½ðnþ 1Þ=2�ðL=2nÞ ¼Lðnþ 1Þ=ð4nÞ voters prefer a lower tariff than the equilibrium. This same
expression also gives a lower limit on the number of voters who prefer a higher
tariff. For the case with n even, at leastP
j¼1;n=2 ðL=2nÞ ¼ L=4 voters prefer a lowertariff, and at least that many prefer a higher tariff, than the equilibrium. Thus,
although we know that the equilibrium tariff may differ from the median voter’s
optimum, at least we know that it cannot be lower than the 25th percentile voter’s
optimum, or higher than the 75th percentile voter’s.
Second, in a well-defined sense these bounds are minimal. Consider an arbitrary
distribution F and let n be odd. It is easy to see that we can divide up the population
among the n districts in such a way that the equilibrium tariff comes arbitrarily
close to the 25th percentile voter’s optimum. Give each district an index number j
from 1 to n and normalize the population size to unity. Take the richest half of the
population and divide it evenly among the n districts. (It does not matter exactly
which voters within this set are allocated to which districts.) Now, for some small
positive 1; move a mass ½ðnþ 1Þ=ðn2 1Þ�1 of voters into each district with an
index j strictly greater than ðnþ 1Þ=2 and move 1 out of each district with an indexless than or equal to ðnþ 1Þ=2: Now, each district with an index j between 1 and
ðnþ 1Þ=2 has amass of ½1=ð2nÞ2 1� voters. Fill out district 1 by adding the poorest½1=ð2nÞ þ 1� voters, then fill out district 2 by adding the poorest ½1=ð2nÞ þ 1� votersnot yet allocated, and so on until all districts have been filled (and hence all voters
have been allocated). Each district 1 through ðnþ 1Þ=2 will have a median voter
drawn from the bottom half of the population, with the median voter for district
jþ 1 richer than that from j; while each district with a higher index will have
a median voter drawn from the top half of the population. Therefore, district
ðnþ 1Þ=2 is the median district. In addition, the median voter of that district will be
richer than ½ðn2 1Þ=2�½1=ð2nÞ þ 1� þ 1=ð2nÞ ¼ ðnþ 1Þ=ð4nÞ þ ½ðn2 1Þ=2�1voters. By appropriate choice of 1; this can be made as close to ðnþ 1Þ=ð4nÞ asone wants. The even case and the case for the upper limits are parallel.
This is summarized in Proposition 3.
Proposition 3. Fix a national wealth distribution, F:
ðiÞ In the case n ¼ 1; the unique equilibrium tariff outcome is the most preferred
tariff of the median voter.
Trade Policy Making by an Assembly 21
(ii) In the case n . 1 with n even, the least upper bound of equilibrium tariff
outcomes sustainable by appropriate allocation of voters to districts is
the most preferred tariff of the 75th percentile voter, and the greatest lower
bound is the most preferred tariff of the 25th percentile voter in the case
in which n is even. In the odd case, the limits are ðnþ 1Þ=ð4nÞ and ½12
ðnþ 1Þ=ð4nÞ�; respectively.
It is, then, clear that the Mayer n ¼ 1 case is really quite special. This is the only
case in which the set of equilibria is a point. Indeed, the case n ¼ 2 looks much
more like the case n ¼ 500 than like the case n ¼ 1; both of the latter cases have
the same range of possible outcomes. Clearly, since the parliamentary model is
much more like what real-world political institutions look like, this suggests that
empirical work needs to address somehow the importance of inter-regional, intra-
national distribution of wealth for the determination of trade policy.
One difference that this implies from the Mayer model involves the
international pattern of trade policies. The Mayer model predicts that if in each
country median income is below mean income, which is generally the case in
practice, all capital-abundant (rich) countries should use a tariff and all labor-
abundant (poor) countries should have free trade or subsidize their imports. Of
course, this is empirically absurd; all countries, rich or poor, have historically had
positive protection rates with very few exceptions. However, the assembly model
makes no such prediction. For example, if allocations of voters to districts are such
that in rich countries the 30th percentile voter is the one whose optimal tariff is
implemented, but in poor countries it is the 65th percentile, it is quite possible that
the tariff will be the most preferred of a below-average-income voter in the rich
country and an above-average-income voter in the poor countries, leading to
positive tariffs everywhere. Whether or not this is the case in practice is, of course,
a tricky empirical question.
The empirical implementation of the Mayer model by Dutt and Mitra (2002)
suggests a convenient way of summarizing the differences between the unitary
government case and the assembly case. They summarize the empirical
implications of the n ¼ 1 model with an estimating equation whose essence can
be summarized (if over-simplified) as follows
ti ¼ aþ bð �Yi 2 Ymedi Þ þ g ð �Yi 2 Y
medi Þð �YiÞ þ 1i;
where ti denotes the average tariff recorded for country i; Ymedi is the median
income in country i; and �Yi denotes the average per capita income for country i;while 1i; is a disturbance term. Dutt and Mitra use this as a regression equation to
explain differences in levels of protection across countries. The explanation is
J. McLaren and B. Karabay22
that, under the Mayer model, ð �Yi 2 Ymedi Þ is a measure of the political distortion
away from free trade in country i (since if this is equal to zero the equilibrium is
free trade), but the direction in which that distortion acts depends on the overall
level of capital per worker in that country. In rich countries, rich voters want free
trade and vice versa in poor countries, so the sign of the interaction term gmust be
positive. Dutt and Mitra (2002) show that this is supported in the data. In the
assembly model a similar logic is valid, but the political distortion is captured by
the term ð �Yi 2 Ymed;medi Þ; where Ymed;med
i indicates the income of the median voter
in the median district, as discussed above.
Two ways of writing this political distortion variable can help clarify the
relationship between the Mayer model and the assembly model. The first is
ð �Yi 2 Ymed;medi Þ ¼ ð �Yi 2 Y
medi Þ þ ðY
medi 2 Y
med;medi Þ
The first term is the explanatory variable used in Dutt and Mitra (2002), and
measures the difference between the mean and median income in country i: Thesecond term measures the difference between the population median and the
median of the median district. Under the assembly model, this would be an
omitted variable that could bias the Dutt and Mitra regression. Clearly, if all
districts have the same distribution of wealth, the second term can be ignored and
the Mayer model will predict well, so one interpretation is that the omitted
variable is a measure of inter-district heterogeneity. The second way of writing the
political distortion term is
ð �Yi 2 Ymed;medi Þ ¼ ð �Yi 2 �Y
medi Þ þ ð �Y
medi 2 Y
med;medi Þ;
where �Ymedi is the average income in the median district. The first term measures
the difference between average income in the median district and average income
countrywide, and the second term measures the difference between mean and
median income in the median district. One can take the first term to be a measure
of inter-regional income inequality and the second to be a measure of intra-
regional income inequality (or skew, more precisely). An interesting empirical
question is: which is more important determinant of trade protection in practice?
Special cases. A handful of special cases of the inter-district distribution of
voters illustrate the behavior of the assembly model. First, clearly, if all districts
are identical ðFj ¼ F ;jÞ; the predictions are exactly as for the Mayer model.
Second, if the districts are completely different, so that the support of Fj and the
support of Fk do not intersect for j – k; then if n is large enough the tariff under
the assembly model will closely approximate the tariff under the Mayer model.
The reason is that the national median voter will, in that case, necessarily be
contained in the median district (recalling that all districts contains the same
Trade Policy Making by an Assembly 23
population mass), and with n large the range of support for the median district will
be small. Therefore, the difference between the median of the median district and
the national median will be small.
These two examples illustrate the point that heterogeneity of districts is a
necessary, but not a sufficient, condition, for the predictions of the Mayer and
assembly models to differ.
A final special case is a parallel to the special case discussed at the end of the
specific factors model. Fix the aggregate endowments L of labor and K of capital,
and let n be odd for concreteness (of course, L is also the number of voters).
Suppose that there are two kinds of voter: Those with no capital (“workers”) and
those with kp units of capital (“capitalists”). In addition, there are two kinds of
districts: there are m districts that have some workers and some capitalists, and
there are n2 m districts that have only workers. For all the mixed districts, the
number of capitalists in the district is equal to K=ðmkpÞ: As m ranges from 1 to n;the geographic distribution of capitalists becomes less concentrated and the
number of capitalists in each mixed district falls. Clearly, if m is less than
ðnþ 1Þ=2, there will be a positive tariff (because the median voter of the median
district will be a worker with no capital). At the same time, if K=ðnkpÞ . L=ð2nÞ;or K=L . kp=2 and m $ ðnþ 1Þ=2; there will be free trade in equilibrium,
because even if the capitalists are spread as thinly as possible with m ¼ n; themedian voter in each district will be a capitalist. If 2K=ððnþ 1ÞkpÞ , L=ð2nÞ; orK=L , kpðnþ 1Þ=4n; there will be a positive tariff regardless of m; since even if
the capitalists are as concentrated as possible subject to the constraint that the
median district be mixed (in other words, even if m ¼ ðnþ 1Þ=2Þ; the median
voter in mixed districts will be a worker. Finally, if kpðnþ 1Þ=4n , K=L , kp=2;then if m , ðnþ 1Þ=2; there will be a tariff; if m ¼ n; there will be a tariff; and fora range for m in between these extremes, free trade will obtain in equilibrium.
To summarize, for this special case with homogeneous workers and
homogeneous capitalist, if capitalists are found in a majority of districts, then
if there is a large enough endowment of capital, capitalists will have their preferred
trade policy; if the endowment is very small, the workers will have their
preferred trade policy; and if there is a moderate endowment of capital, then
capitalists will have their preferred trade policy only if “m” is in a middle range, or
only if the capitalists are moderately geographically concentrated.
2.4. THE ELECTORAL COLLEGE
As a side benefit to pursuing this assembly model, it is easy to see that it is almost
isomorphic to a straightforward model of presidential electioneering under the
J. McLaren and B. Karabay24
electoral college system of the United States. Under that system, each state
receives a certain number of “electoral college votes”, based on the state’s
population. The candidate who receives the largest number of ballots cast by
voters in a state receives all of that state’s votes in the electoral college; the
candidate with the largest number of electoral college votes is named president.
To keep the argument simple, assume (in contrast to previous sections) that the
president has sole decision-making power over trade policy. For example, this
could be the case if the presidential veto power gives all of the bargaining power to
the president in his dealings with congress. Of course, for the US case, both the
models in the previous sections with no president at all, and the present model
with a president possessing unchecked powers, are extremes posited in order to
focus on one issue at a time. A point that will emerge in this discussion is how
similar the behavior of the two models are, despite their polar-opposite
assumptions, once the effect of the electoral college is taken into account.
For concreteness, let us adapt the Mayer model of Section 2.3. If there are two
nationwide candidates who campaign by making credible nationwide commit-
ments to subsequent trade policy, then the unique equilibrium tariff policy will
be characterized as follows. Denote the number of electoral college votes of state
j by wj; and assume that the wj values are proportional to the number of voters
in each state. Order the states from 1 to 50 by the most preferred tariff of
the median voter of each state. Define the cumulative electoral votes of state J
by CðJÞ ;P
j¼1;J ½wj�: Define the weighted-median state Jp by CðJp 2 1Þ ,
Cð50Þ=2 and CðJpÞ . Cð50Þ=2: This exists unless a state J 0 has CðJ 0Þ ¼ Cð50Þ=2exactly, which is, of course, unlikely in practice.
Then, if the weighted-median state exists, the unique equilibrium outcome
is the most preferred tariff of that state’s median voter. If no weighted-median
state exists, any tariff between the most preferred tariff of state J 0 just defined
and that of state J0 þ 1 is an equilibrium outcome.
Thus, we are back to the same outcome as under the assembly model, the
median-of-medians rule, with the exception that the states are not of equal size.
We have the same 25–75 bounding rule as before (the realized tariff cannot be
below the most preferred level of the 25th-percentile voter, or above that of the
75th-percentile voter), although these are not tight bounds in the greatest-lower-
bound and least-upper-bound sense of the assembly model (Proposition 3),
owing to the differences in size between states. (Clearly, if one state has 99% of
the population in it, that state will always be the weighted-median state, and the
equilibrium tariff will always be between the 49th and 51st percentiles of the
national distribution.) With that one qualification, the analysis of the assembly
case applies to this case.
Trade Policy Making by an Assembly 25
It should perhaps not be surprising that the logic of an elected assembly applies
to the case of an electoral-college system, since originally the electoral college was
a real elected assembly, elected with the power to choose the president through due
deliberation. However, the point seems to be underappreciated in the economic
literature on this subject. This is surprising, since casual evidence of the effects of
the electoral college system on US trade policy appears to be abundant. A sharp
example is the 1888 presidential election that brought Benjamin Harrison to
power.1 In that election, the free-trader Grover Cleveland won the popular vote but
the protectionist Harrison won in the electoral college. Tariffs were a major issue in
the campaign, with Harrison promising to raise them and Cleveland proposing to
lower them. Thus, a model of the median-voter type would have predicted a liberal
trade regime, but instead the country received a protectionist president who signed
the notorious McKinley tariff into law 2 years later. More recently, the 2002
executive order to levy tariffs on a variety of steel products appears to be motivated
by the fact that Pennsylvania is a swing state in presidential elections with a
generous number of electoral college votes, and Pennsylvania steel workers could
be expected to feel grateful for the protection. The relationship between the
electoral college and US trade policy appears to be ripe for research.
2.5. CONCLUSIONS AND OPEN QUESTIONS
The principal conclusions can be summarized as follows.
1. There is a large difference between the outcomes predicted by a unitary model
and by a model of government by assembly. In the simple, standard version
studied here, the former predicts a median-voter rule (the most-preferred tariff
of the national median voter is implemented), while the latter predicts a
median-of-medians rule in which the most-preferred tariff of the median voter
of the median district is implemented. This can result in large differences in the
level of the tariff.
2. Perhaps surprisingly, the range of possible equilibria does not depend on the
number of districts, as long as there are more than one. Under the unitary model,
for a given distribution of trade policy preferences, the range of possible
outcomes is a singleton (namely, the 50th percentile most-preferred tariff).
Under the assemblymodel, the range is from the 25th to the 75th percentilemost-
preferred tariff, depending on how voters are divided up between districts. Thus,
there is a large qualitative difference between the unitary and assembly models,
even if there are as few as two districts.
1 We are grateful to Doug Irwin for pointing out this example to us.
J. McLaren and B. Karabay26
3. Ceteris paribus, we are more likely to see protection the more abundant are
factors intensive in import-competing sectors (subject, of course, to the
constraint that they are not so abundant that they become export sectors).
4. On the other hand, the likelihood of protection is non-monotonically related to
within-country geographic concentration of import-competing factors. If they
are too concentrated, they can never command amajority in parliament, while if
they are too disperse, they will not dominate any one district, and thus will
control no seats in parliament.
5. Thus, not only national income distribution should matter for trade policy (as in
Mayer (1984) and Dutt and Mitra (2002)), but within-country inter-regional
distribution of income should matter as well.
6. The electoral college system of presidential elections has similar properties to
the government-by-assembly model.
This all raises a number of empirical questions far beyond the scope of this
chapter.
1. In the specific-factors model, the degree of concentration of the import-
competing industry across the country was important for determining the
degree of protection obtained, a variable that has no role in standard models of
unitary government. How much explanatory power does it have for tariff levels
internationally?
2. In the Heckscher–Ohlin–Mayer model, the degree of concentration of the
import-competing-intensive factor across the country was important for
determining the degree of protection obtained. The same question arises as in
the previous point, with the further question of how the geographic industry
concentration compares with geographic factor concentration as an explanatory
variable.
3. The assembly model also makes clear that within-country inter-regional
income inequality and intra-regional income inequality have separate
important roles in tariff determination (see the discussion following
Proposition 3). What is the relative importance of these two variables in
the determination of trade policy?
4. One of the great perplexing features of modern trade policy is the propensity
of rich countries to protect their agricultural sectors, despite the fairly small
minority who derive their living from the sector. Is it possible to explain this
using the assembly model, without recourse to treating agriculture as an
organized interest group? In other words, is the control of representatives
from farm states a valuable enough prize to risk antagonizing all other voters
by maintaining agricultural protection?
Trade Policy Making by an Assembly 27
More generally, we have seen that the assembly model shows one mechanism
by which the power of a minority can be magnified (in that as small a group as
one-quarter of the population can receive its desired policy at the expense of the
majority). This is a mechanism completely separate from the more familiar
mechanism of organized interest groups. It would be desirable to identify which of
these two mechanisms is more important in practice, but how to do that is not
obvious.
Finally, we can speculate about additions to the model that have the potential to
move the equilibrium even farther away from the median voter. First, it seems
natural to add a non-policy element to voters’ preferences, as in Lindbeck and
Weibull (1993). Suppose that each voter has a preference for one candidate over
another even if they commit to the same policy, because of charisma, ethnic
identity, or some other exogenous reason. Then, if these non-policy preferences
are correlated for voters within a district, it is possible, for example, that the
median district will be so biased in favor of one of the parties that neither party
enters that district into its calculations of its optima policy stand: that district is
assumed to be locked-up for one of the parties no matter what policies are
announced. In that case, the equilibrium policy will not be the most preferred of
the median voter of the median district, and could lie outside of the 25–75 bounds
established for this model. Second, in some legislatures committees that draft
legislation seem to be very influential (see Krehbiel (1991) for an exhaustive
study). To the degree that a committee structure confers disproportionate power
over legislation on a small number of members of the assembly, it seems
reasonable to ask if they might permit movements of policy farther away from the
median. Both of these extensions are, however, beyond the scope of this chapter.
ACKNOWLEDGEMENTS
The authors are grateful to Erica Gould and Douglas Irwin for helpful suggestions.
All errors are our own responsibility.
REFERENCES
Baily, M. and Brady, D. W. (1998). Heterogeneity and representation: the senate and free
trade. American Journal of Political Science, 42, 524–544.
Bailey, M., Goldstein, J. and Weingast, B. R. (1997). The institutional roots of American
trade policy: politics, coalitions and international trade. World Politics, 49, 309–338.
J. McLaren and B. Karabay28
Baldwin, R. and Magee, C. (2000). Is Trade policy for sale? Congressional voting on recent
trade bills. Public Choice, 105, 79–101.
Dennis, C., Bishin, B. and Nicolaou, P. (2000). Constituent diversity and congress: the case
of NAFTA. Journal of Socio-Economics, 29, 349–360.
Dutt, P. and Mitra, D. (2002). Endogenous trade policy through majority voting: an
empirical investigation. Journal of International Economics, 58, 107–134.
Irwin, D. L. and Krozner, R. S. (1999). Interests, institutions, and ideology in securing
policy change: the republican conversion to trade liberalization after Smoot-Hawley.
Journal of Law and Economics, 42, 643–673.
Findlay, R. and Wellisz, S. (1982). “Endogenous tariffs, the political economy of trade
restrictions, and welfare,” in Import Competition and Response, J. Bhagwati (eds.),
Chicago: University of Chicago Press.
Grossman, G. and Helpman, E. (1994). Protection for sale. American Economic Review, 84,
835–850.
Krehbiel, K. (1991). Information and Legislative Organization, Ann Arbor, MI: University
of Michigan Press.
Lindbeck, A. and Weibull, J. W. (1993). A model of political equilibrium in a
representative democracy. Journal of Public Economics, 51, 195–209.
Lohmann, S. and O’Halloran, S. (1994). Divided government and U.S. trade policy: theory
and evidence. International Organization, 48, 595–632.
Mayer, W. (1984). Endogenous tariff formation. American Economic Review, 74, 970–985.
Milner, H. V. (1999). The political economy of international trade. Annual Review of
Political Science, 2, 91–114.
Nelson, D. R. (2002). The Political Economy of Trade Policy Reform: Social Complexity
and Methodological Pluralism, Working Paper, Tulane University.
Peltzman, S. (1985). An economic interpretation of the history of congressional voting in
the twentieth century. American Economic Review, 75, 656–675.
Poole, K. T. and Rosenthal, H. (1997). Congress: A Political-Economic History of Roll Call
Voting, New York: Oxford University Press.
Putnam, R. (1988). Diplomacy and domestic politics: the logic of two-level games.
International Organization, 42, 427–460.
Trade Policy Making by an Assembly 29
CHAPTER 3
Should Policy Makers be Concerned
About Adjustment Costs?
CARL DAVIDSONa,b,* and STEVEN J. MATUSZa,b
aDepartment of Economics, Michigan State University, East Lansing, MI 48824, USAbGEP, University of Nottingham, Nottingham, UK
Abstract
Using a general equilibrium model that incorporates labor market turnover, we
estimate the adjustment costs associated with trade reform. Our key finding is that
estimates of these costs increase significantly when the training and job-acquisition
processes are modeled. For example, we estimate that adjustment costs eat away
roughly 10–15% of the benefits from liberalization when we include the time cost
of retraining. When we also include the resource costs of retraining, our estimates
jump to 30–80%. These results contrast sharply with earlier estimates that
indicated that adjustment costs are likely to be small relative to the gains from trade.
Keywords: Adjustment costs, liberalization, training, search
JEL classifications: F1, F16
3.1. INTRODUCTION
Even the most strident advocates of free trade would readily admit that it takes
time for economies to reap the benefits from trade liberalization. As trade patterns
*Corresponding author. Address: Department of Economics, Michigan State University, East Lansing,
MI 48824, USA.
E-mail address: [email protected]
change, some workers lose their jobs and must seek reemployment in expanding
sectors. There may be some cases in which these workers need to retool in order
find new jobs. Of course, these workers do not produce any output while they
search for reemployment and/or retrain. As a result, during the adjustment
process, there may be a period during which welfare falls below its initial level.
Policy makers often have a difficult time weighing these short run adjustment
costs against the long run benefits from freer trade and this has made some
countries reluctant to reduce barriers to trade. Growing concern about the
importance of the adjustment process in the policy community is evident, as recent
studies commissioned by the World Bank (Brahmbhatt, 1997) and the WTO
(Bacchetta and Jansen, 2003) clearly indicate. Beyond concerns for equity, a full
understanding of the magnitude and scope of adjustment can inform our views of
the political economy of protectionism. In this chapter, we investigate the nature
of the adjustment process and try and get some handle on the magnitude of the
costs involved in order to determine whether such concerns are warranted.
Recent research suggests that the personal cost of worker dislocation may be
quite high. Jacobson et al. (1993a,b) find that the average dislocated worker
suffers a loss in lifetime earnings of $80,000. Yet, as disturbing as this finding may
be, it tells us nothing about the aggregate costs of adjustment. It is quite possible
for individual workers to lose a great deal while at the same time the economy is
suffering only minor aggregate adjustment costs. Nevertheless, those who oppose
trade liberalization often point to such personal losses, along with wage losses to
those who remain employed in import competing industries, and ask whether the
gains from freer trade are really worth such costs. Academic economists tend to
dismiss such concerns by either suggesting that the aggregate costs of adjustment
are probably very small compared to the gains from trade or by pointing out that
the gains from trade are always large enough that we can fully compensate all
those who suffer personal losses without exhausting the gains. Unfortunately,
there are problems with both of these arguments. The latter argument ignores the
fact that such compensation rarely, if ever, takes place. And, the problem with the
former argument is that there is almost no solid research on which to base such
claims. In other words, we know very little about the magnitude and scope of
aggregate adjustment costs.
Estimates of aggregate adjustment costs are rare.1 The two main contributions
are Magee (1972) and Baldwin et al. (1980), both of which follow a similar
approach. First, estimates were made about the number of workers who would
1 A number of authors have attempted to measure adjustment costs within specific industries. See, for
example, de Melo and Tarr (1990) who focus on the US textile, auto and steel industries or Takacs and
Winters (1991) who studied the British footwear industry.
C. Davidson and S.J. Matusz32
lose their jobs due to liberalization. These job losses were then evaluated based on
an appropriate measure of the displaced workers’ wages. Finally, the authors then
assumed that these workers would find reemployment after a length of time
determined by estimates of the average duration of unemployment. Both papers
conclude that adjustment costs are probably very small when compared to the
gains from liberalization. For example, with a 10% discount rate, they both
estimate that the short run costs of adjustment would eat away no more than 5% of
the long run gains from trade.2
It is hard to know what to make of these estimates. Neither paper attempts to
take into account either the time or resource costs that are involved in the
retraining that dislocated workers may be forced to go through. The resource cost
of job search is also ignored. Moreover, since the reemployment process is not
modeled, it is hard to take into account any displacement that may occur as
dislocated workers find reemployment in new sectors. There are other problems as
well, but all stem from the same basic issue—since there is no model of the
adjustment process underlying these estimates, there may be many general
equilibrium spillover effects that are not being captured. This is not intended as a
criticism of these papers. At the time that these papers were written, rigorous
models that explicitly allow for the trade frictions and informational asymmetries
that lead to equilibrium unemployment were only in their infancy.3 It would have
been difficult to extend the type of general equilibrium models typically used for
trade analysis to allow for equilibrium unemployment and retraining. The
empirical approach adopted by these authors was entirely appropriate given the
state of the trade literature at that time.
There is another line of research that is relevant for what follows. This research
simply assumes that adjustment costs are non-trivial and then examines the
implications of this for the optimal time path of liberalization. It has been argued
that in order to minimize adjustment costs, trade barriers should be removed
gradually (see, for example, Cassing and Ochs, 1978; Karp and Paul, 1994;
Gaisford and Leger, 2000; Davidson and Matusz, 2004).4 The rationale behind
2 In a recent paper, Trefler (2001) examined the short-run adjustment costs and long-run efficiency
gains from trade liberalization in Canada by quantifying the impact of NAFTA on specific
Canadian industries subject to large tariff cuts. He found evidence of substantial short-run
adjustment costs—a 15% decline in employment and a 10% decline in output. Balanced against
these costs were only moderate productivity gains of about 1% per year. However, Trefler made
no attempt to measure the aggregate gains and losses from NAFTA.3 We are referring to the literatures on trading frictions (search theory), efficiency wages, and
insider/outsider models of the labor market, among others.4 For a survey of this literature, see Falvey and Kim (1992). Other recent contributions that do not focus
on the role of congestion include Li and Meyer (1996) and Furusawa and Lai (1999).
Should Policy Makers be Concerned About Adjustment Costs? 33
this rests on the assumption that as workers flee the import-competing sector and
seek jobs in the expanding export sectors, congestion externalities will arise that
increase the cost of adjustment. If the government removes trade barriers slowly, it
can control the flow of workers, reduce congestion, smooth out the adjustment
process and minimize the social cost of adjustment.
Our goal in this chapter is to build on recent advances in the theory of
equilibrium unemployment by presenting a simple general equilibrium model of
trade that includes unemployment and training. We then use the model to explore
the scope and magnitude of adjustment costs relative to the gains from trade. In
our model, workers differ in ability and jobs differ in the types of skills they
require. Workers sort themselves by choosing occupations based on expected
lifetime income. These workers then cycle between periods of employment,
unemployment and training with the length of each labor market state determined
by the turnover rates in each sector. One of the advantages of the model is that it is
simple enough to allow us to solve analytically for the adjustment path between
steady states thereby allowing us to calculate the adjustment costs associated with
trade reform. Another advantage is that many of the key parameters (e.g., labor
market turnover rates) are observable so that we can rely on existing data to
determine their likely values. However, one of the shortcomings of the model is
that there are few existing estimates on which to base assumptions about the
resource and time costs associated with training and these values play important
roles in our analysis. We, therefore, solve the model for a wide variety of
assumptions about these values and look for conclusions that are robust.
In developing our model, we purposefully abstract from congestion
externalities by assuming that after liberalization, the job acquisition rate in the
export sector remains at its pre-liberalization level. We do so for two reasons: first,
our main goal is to show that by including the resource and time costs associated
with training and job search we obtain estimates of adjustment costs that are
substantially larger than those in the existing literature. In doing so, we want to
ensure that our estimates are conservative, and by abstracting from congestion
externalities, we are likely to be underestimating the true magnitude of these costs.
Moreover, this ensures that our estimates are not driven by a (potentially)
controversial assumption about the search process. The second reason that we
assume away congestion has to do with the nature of the liberalization process
itself. In this chapter we are not interested in adding to the literature on the optimal
time path of liberalization. As we mentioned above, previous work has shown that
when congestion externalities are present, removing trade barriers gradually can
lower aggregate adjustment costs. However, in the absence of congestion, there is
no reason for gradualism. By assuming away congestion externalities we can keep
C. Davidson and S.J. Matusz34
our analysis simple and focus on the magnitude of the adjustment costs that arise
when liberalization is complete and immediate.
Our results are surprising and contrast sharply with the previous literature.
First, our model predicts that adjustment will take place relatively quickly, with
net output returning to its pre-liberalization level within 2.5 years. This result,
which is partly due to our assumption that post-liberalization export sector labor
markets are not troubled by congestion, implies that an empirical analysis of
adjustment based on yearly data could easily lead to the conclusion that
adjustment costs are quite small. However, this is not the case in our model. Even
with our most modest assumption concerning training costs we find that their
inclusion in the model significantly increases our estimates of aggregate
adjustment costs. For example, we find that when we take the time cost of
retraining into account the short run adjustment costs amount to (at least) 10–15%
of the long run benefits from liberalization. When the resource costs of retraining
are taken into account as well, our estimates jump to 30–80% of the long run
gains from freer trade!5 The fact that we obtain these results in a model in which
the job acquisition rate in the export sector does not fall after liberalization is
particularly noteworthy.
In the latter part of the chapter, we turn to a related issue, and ask whether
there is any way to know a priori which types of economies are likely to face
relatively large adjustment costs. Labor markets and the institutions that govern
them vary greatly across the world. Jobs tend to last longer in the US than they
do in Europe and Japan. The average duration of unemployment is relatively
short in the US, while it can be quite long in some European countries. The
implication is that all labor market turnover rates tend to be higher in the US
than they are in most European countries. In addition, wages are more flexible
in US labor markets than they are in their European counterparts. Consequently,
labor economists typically characterize US labor markets as flexible while
European labor markets are considered sluggish. One would expect that the
flexibility of the labor market would play a key role in determining the relative
importance of adjustment costs.
We investigate this issue by determining how the ratio of adjustment costs to
the gains from trade varies as turnover rates increase uniformly. In our model,
we find, perhaps as expected, that relative adjustment costs are decreasing in the
degree of labor market flexibility so that economies with slothful labor markets
face higher costs of adjustment than economies with either flexible or sluggish
5 It is worth noting that while our conclusion that short-run adjustment costs may be quite high is quite
different from those reached byMagee (1972) andBaldwin et al. (1980), it is consistentwith basicmessage
of Trefler’s (2001) study of the impact of NAFTA on select Canadian industries (see Footnote 2).
Should Policy Makers be Concerned About Adjustment Costs? 35
labor markets. However, somewhat surprisingly, we find that the net benefits
from trade reform have the same relationship with labor market flexibility so
that economies with slothful labor markets have the most to gain from
liberalization.
This surprising result has its roots in the manner in which tariffs distort
economies with different degrees of labor market flexibility. We find that tariffs
distort slothful labor markets more than sluggish ones. The removal of the tariff,
therefore, generates large benefits in such economies; in fact, they are even large
enough to swamp the economy’s high level of short run adjustment costs. As a
result, economies with the most slothful labor markets gain the most from trade
liberalization.
In the conclusion, we discuss the appropriate way to view our results. We
emphasize that although our estimates of adjustment costs are quite high, this
should not be misinterpreted as a warning about the dangers of liberalization.
Instead, we argue that economists and politicians should spend more time
worrying about the appropriate way to compensate those who bear the burden
of these costs and that these policies should be an important component of the
liberalization process. We also point out that our results suggest that the cost
of new protectionist policies may be substantially higher than previous
estimates indicate since newly created barriers to trade generate adjustment
costs as well.
3.2. THE MODEL
3.2.1. Background
In developing our model, we have several goals in mind. First, we want to use a
general equilibrium trade model that is rich enough to capture some essential
features of the employment process. In particular, we want a model that explicitly
allows for both a training process in which workers acquire the skills required to
find a job and a search process that those same workers must go through to find an
employer. Second, we want to keep the model simple and tractable in order to be
able to solve analytically for the transition path between steady states. This allows
us to calculate the adjustment costs associated with trade reform. Third, we want
the model to be general enough to allow for cross-country differences in labor
market structure so that we may investigate the relationship between labor market
flexibility and adjustment costs.
The basic structure of the model is as follows. We have an economy in which
workers with differing abilities must choose between two types of jobs—those that
C. Davidson and S.J. Matusz36
do not require many skills and offer low pay and those that require significant
training and pay relatively high wages. Jobs in the low-tech sector are easy to find,
do not last very long (there is high turnover) and require skills that are job specific.
In contrast, high-tech jobs are relatively hard to find, presumably because the
matching problem is harder to solve, last longer once employment is secured and
require a combination of job specific and general skills. We assume that in each
sector high-ability workers produce more output than their low-ability
counterparts. Under certain assumptions, this implies that in equilibrium workers
sort themselves so that high-ability workers train for high-tech jobs while low-
ability workers are drawn to the low-tech sector.
We begin by assuming that the low-tech sector is protected by a tariff. This
raises the return to training in that sector and causes some workers who should
train in the high-tech sector to seek low-tech jobs instead. When the tariff is
removed, these workers shift to the high-tech sector. This shift is gradual,
however, since these workers will first have to enter the high-tech training process
and then search for jobs. In addition, some of the workers who may eventually
want to shift sectors may already hold low-tech jobs and since training and search
are costly, they may choose to wait until they lose their low-tech jobs before
making the switch. As a result, it may take significant time before the economy
gets close to the new steady state. In this setting, adjustment costs are measured by
comparing what the economy could gain if it could jump immediately to the new
steady state with what it actually gains taking into account the costly transition
that it experiences in moving to the new steady state.
3.2.2. Formalizing the model and finding the initial steady state
We consider a continuous time model of a small open economy consisting of two
sectors and a single factor of production, labor. We use ai to denote worker i’s
ability level and we assume that ai is uniformly distributed across ½0; 1� with the
total measure of workers equaling L: To obtain a job in either sector, workers must
first acquire the requisite skills. Training is costly, both in time and resources. In
sector j;workers seeking a job must pay a flow cost of pjcj while training, where pjdenotes the price of good j (so that sector j training costs are measured in units of
the sector j good). The length of the training process is assumed to be random,
with sector j trainees exiting at rate tj: This implies that the average length of
training in sector j is 1=tj: Our notion that training is more costly both in time and
resources in sector 2, the high-tech sector, is captured by assuming that c1 and t2are small while c2 and t1 are large (we will be more precise below). We use LjTðtÞ
to denote the measure of workers training in sector j at time t:
Should Policy Makers be Concerned About Adjustment Costs? 37
After exiting the training process, workers must search for employment.6 Jobs
in the low-tech sector are plentiful, so that jobs are found immediately.7 In
contrast, it takes time to find high-tech jobs and we use e to denote the steady-state
job acquisition rate in that sector. It follows that in the initial steady state the
average spell of sector 2 unemployment is 1=e: We use LSðtÞ to denote the
measure of workers searching for high-tech jobs at time t:Once a job is found, a type i sector j worker produces a flow of qjai units of
output as long as he/she remains employed. Since output is increasing in ai; higherability workers produce more than their lower ability counterparts in each sector.8
This output is sold at pj and all of the revenue goes to the worker in the form of
earned income (so that the sector j wage earned by a type i worker is pjqjai). We
assume that in a steady-state sector j workers lose their jobs at rate bj; so that the
average duration of a sector j job is 1=bj: Since high-tech jobs are assumed to be
more durable than low-tech jobs, it follows that b1 . b2: The measure of workers
employed in sector j at time t is denoted by LjEðtÞ:Upon separation, a worker must retrain if his/her skills are job specific. In
contrast, if his/her skills are general, he/she can immediately begin to search for
reemployment. As noted above, we assume that the skills acquired during low-
tech training are job specific. We make this assumption because, to us, it seems
natural. While training, a store clerk may need to learn the layout of the store in
which he/she is employed, the procedure involved in opening and closing the
store, the functioning of a particular type of cash register, and so on; but, in
gaining this knowledge the worker learns nothing about how to prepare fast food
(or perform other low-skill tasks). In contrast, high-tech workers like accountants,
managers and lawyers all must complete college and obtain some post-graduate
education. If they lose their job, many of these workers will be able to obtain
reemployment in the same field and in doing so they will not be required to go
back through school. Moreover, even if these workers choose to change
occupation, they will have acquired some general skills along the way that may
6 The assumption that the training process takes place before search is not crucial for the analysis.
We could assume instead that training takes place after completion of search without altering the nature
of our results.7 Of course, many low-ability workers face difficulty finding any job whatsoever and therefore
face a long expected duration of unemployment whenever they lose their job. We believe that this
is largely due to their work history and overall ability level. By assuming that low-tech jobs are
plentiful (so that sector 1 employment can be found immediately), we are trying to capture the
notion that the marginal worker (who has the ability to train for a high-tech job) would be able to
find menial employment quite easily if she chooses to do so.8 Ability could refer to attributes that the worker is born with, or it could refer to a combination
of attributes that are either innate or acquired during the elementary education process.
C. Davidson and S.J. Matusz38
allow them to land new jobs without acquiring additional skills. The implication
is that all unemployed low-tech workers need to retrain in order to find
reemployment, while some high-tech workers can move into a new job without
having to retrain. To make this precise, we assume that with probability f high-
tech workers need not retrain after losing their jobs.
The dynamics of the two labor markets are depicted in Figures 3.1 and 3.2.
The evolution of labor markets over time can be described with the aid of these
figures. Let _XðtÞ denote the growth rate of X at time t: These growth rates can be
found by comparing the flows into and out of each labor market state. For
example, in sector 1, the flow out of training is equal to the measure of workers
who complete the training process and take low-tech jobs, t1L1TðtÞ: The flow into
training is equal to the measure of low-tech workers who lose their jobs due to
exogenous separation, b1L1EðtÞ: It follows that the growth rate of low-tech traineesis given by
_L1TðtÞ ¼ b1L1EðtÞ2 t1L1TðtÞ ð1Þ
Similar logic can be used to find the growth rates of employment, _L2EðtÞ; and the
unemployment pool in sector 2, _LSðtÞ: We have
_L2EðtÞ ¼ eLSðtÞ2 b2L2EðtÞ ð2Þ
_LSðtÞ ¼ t2L2TðtÞ þ b2fL2EðtÞ2 eLSðtÞ ð3Þ
In Equation 2, the flow into high-tech employment consists of searching workers
who find employment, eLSðtÞ; while the flow out is made up of employed high-
tech workers who lose their jobs, b2L2EðtÞ: In Equation 3, the flow into the pool of
searchers is made up of those who complete the high-tech training process,
Figure 3.1: Labor market dynamics in sector 1.
Should Policy Makers be Concerned About Adjustment Costs? 39
t2L2TðtÞ; and those workers who lose their high-tech jobs but do not have to retrainbecause their skills are transferable, b2fL2EðtÞ: The flow out of unemployment is
equal to the measure of high-tech searchers who find jobs, eLSðtÞ:9 Finally, in
sector 1, workers are either employed or training, while in sector 2, they are
employed, training, or searching. Thus, we have the following adding up
conditions (where LjðtÞ denotes measure of sector j workers at time t):
L1ðtÞ ¼ L1EðtÞ þ L1TðtÞ ð4Þ
L2ðtÞ ¼ L2EðtÞ þ L2TðtÞ þ LSðtÞ ð5Þ
If we set the left-hand side of Equations 1–3 equal to zero, then we can use
Equations 1–5 to solve for the steady-state measure of workers in each labor
market state. In Appendix A we show that these differential equations can also be
used to solve for the transition path between steady states.
It is important to note that the transition rates (e; b1; and b2) in Equations 1–3
are set at their steady-state values. There are two reasons for this. First, as we
mentioned above, we want to abstract from congestion externalities that might
lower e during the transition period by holding e fixed at its steady-state level.
Second, as we show below, once liberalization occurs all economically inefficient
jobs are immediately destroyed as some low-tech workers quit their jobs and
switch sectors. Thus, the job destruction rate increases endogenously but it does
so instantaneously and then returns to its steady-state level immediately after.
Figure 3.2: Labor market dynamics in sector 2.
9 Similar growth equations for L1E (low-tech employment) and L2T (trainers in sector 2) could also be
defined. However, given the adding up conditions in Equations 4 and 5 they would be redundant.
C. Davidson and S.J. Matusz40
It follows that the increase in job destruction shows up not in the differential
equations describing labor market flows, but in the initial conditions that hold
once liberalization occurs (L2T jumps up immediately once the tariff is removed).
Of course, in order to solve Equations 1–5 we must first explain how to solve
for L1ðtÞ and L2ðtÞ: These values are determined by the behavior of individual
workers, who choose their occupations based on the lifetime income that they
expect to earn in each sector. When workers initially enter the labor market they
have no skills. Thus, their initial choice depends on the relative values of V1T and
V2T; which measure the expected lifetime income for workers training in sectors 1
and 2, respectively. If we define V2S as the expected lifetime income for sector 2
workers who are currently searching for a job and use VjE to denote the expected
lifetime income for employed workers in sector j; then we have the following assetvalue equations (with r denoting the discount rate and g denoting the tariff on
good 1)
rV1TðtÞ ¼ 2p1ð1þ gÞc1 þ t1½V1EðtÞ2 V1TðtÞ� þ _V1TðtÞ ð6Þ
rV1EðtÞ ¼ p1ð1þ gÞq1ai þ b1½V1TðtÞ2 V1EðtÞ� þ _V1EðtÞ ð7Þ
rV2TðtÞ ¼ 2p2c2 þ t2½V2SðtÞ2 V2TðtÞ� þ _V2TðtÞ ð8Þ
rV2SðtÞ ¼ 0þ e½V2EðtÞ2 V2SðtÞ� þ _V2SðtÞ ð9Þ
rV2EðtÞ ¼ p2q2ai þ b2½fV2SðtÞ þ ð12 fÞV2TðtÞ2 V2EðtÞ� þ _V2EðtÞ ð10Þ
In Equations 6–10, the first term on the right-hand side represents flow income.
For employed workers, flow income is equal to the value of the output they
produce ( pjqjai for a type i worker in sector j). Trainees and searching workers
earn nothing while unemployed, and trainees must pay training costs while
acquiring their skills. Thus, current income for searchers is equal to zero while
trainees lose their training costs. The second term on the right-hand side of each
equation is the product of the capital gain (or loss) from changing labor market
status and the rate at which such changes take place. For example, the flow rate
from searching to employment in sector 2 is e while the capital gain associated
with employment is V2E 2 V2S: Note that for workers who are employed in the
high-tech sector, there are two possibilities when they lose their job. With
probability f these workers retain their skills and begin to search for a new job
immediately, while with the remaining probability they must retrain before they
can seek a new job. The final term on the right-hand side, the _V term, represents
Should Policy Makers be Concerned About Adjustment Costs? 41
the growth rate of V : This term captures the appreciation (or depreciation) of the
asset value over time and it is equal to zero in a steady state.
In order to describe the initial steady state equilibrium, we now set each _V term
in Equations 6–10 equal to zero and solve for the expected lifetime income
associated with each labor market state. We obtain (V2S and V2E are omitted since
we do not use them in the subsequent analysis)
V1E ¼p1ð1þ gÞ{ðr þ t1Þq1ai 2 b1c1}
rD1
ð11Þ
V1T ¼p1ð1þ gÞ{t1q1ai 2 ðr þ b1Þc1}
rD1
ð12Þ
V2T ¼p2{t2eq2ai 2 ½ðr þ b2Þðr þ eÞ2 feb2�c2}
rD2
ð13Þ
where D1 ¼ r þ b1 þ t1 and D2 ¼ ðr þ b2Þðr þ t2 þ eÞ þ et2 2 b2fe:Unemployed workers with no skills choose to train in the low-tech sector if
V1T $ max{V2T; 0} and they choose to train in the high-tech sector if V2T $
max{V1T; 0}:Workers with ability levels such that 0 $ max{V1T;V2T} stay out of
the labor market since it is too costly for them to train for any job. These workers
are effectively shut out of the labor market—there are no jobs available for them to
train for since their training costs would exceed any income that they could expect
to earn after finding employment.
As for employed and searching workers, we assume that they are free to change
occupations, but each time they do so they must start out by retraining. It follows
that, in any steady-state equilibrium, these workers never switch sectors.
However, changes in parameters or world prices may result in these workers
changing occupations if the expected lifetime income associated with training in
the other sector exceeds what they expect to earn as a searcher or an employed
worker in their current sector.
To complete the characterization of equilibrium we must place some
restrictions on our parameters. What we have in mind is a model in which
high-ability workers are better suited to produce the high-tech good. It is clear
from Equations 12 and 13 that V1T and V2T are linear and increasing in ai:Moreover, in each sector there is a critical value for ai; denoted by aj; below which
VjTðaiÞ , 0: Workers separate in the desired way if V2T is steeper than V1T at the
initial world prices and if a1 , a2: This is the case if p1ðr þ t1Þq1D2 , p2t2eq2D1
and ðr þ b1Þc1t2eq2 , ½ðr þ b2Þðr þ eÞ2 feb2�c2t1q1: With these two assump-
tions in place, V1T and V2T are as depicted in Figure 3.3. Note that the figure
C. Davidson and S.J. Matusz42
includes two new terms, aL and aH; with aL ; a1 and aH defined as the ability
level for the worker who is just indifferent between training in sector 1 or 2, i.e.,
V2TðaHÞ ¼ V1TðaHÞ:From Figure 3.3, it is clear that workers with ability levels below aL do not enter
the labor force. For these workers, the cost of training for any job is too high.
Workers with ability levels ai [ ½aL; aH� find the low-tech sector more attractive
and choose to train in sector 1. It follows that L1 ¼ ðaH 2 aLÞL: Finally, workerswith ability levels above aH find the high-tech sector relatively more attractive.
These workers train for high-tech jobs, so that L2 ¼ ð12 aHÞL:We can now return to Equations 1–5, set the _L terms equal to zero and solve for
the measure of workers in each labor market state in the initial steady state. We
obtain (LS is omitted since it is not used in the subsequent analysis)
L1T ¼b1ðaH 2 aLÞL
b1 þ t1ð14Þ
L1E ¼t1ðaH 2 aLÞL
b1 þ t1ð15Þ
L2T ¼ð12 fÞeb2ð12 aHÞL
ðeþ b2Þt2 þ ð12 fÞeb2ð16Þ
L2E ¼t2eð12 aHÞL
ðeþ b2Þt2 þ ð12 fÞeb2ð17Þ
Figure 3.3: The equilibrium allocation of workers.
Should Policy Makers be Concerned About Adjustment Costs? 43
where, from Equations 12 and 13,
aH ¼p2c2D1½ðr þ b2Þðr þ eÞ2 feb2�2 p1ð1þ gÞc1D2ðr þ b1Þ
p2t2eq2D1 2 p1ð1þ gÞt1q1D2
ð18Þ
aL ¼ðr þ b1Þc1
t1q1ð19Þ
These values can now be used to determine YSS; the initial steady state value of
output measured at world prices and net of training costs. Since the average low-
tech worker produces 12q1ðaL þ aHÞ units of output while the average high-tech
workers produces 12q2ð1þ aHÞ units, we have
YSS ¼ p1{0:5q1ðaL þ aHÞL1E 2 c1L1T}þ p2{0:5q2ð1þ aHÞL2E 2 c2L2T} ð20Þ
Finally, to turn net output into utility, we assume that all workers have the same
utility function given by UðZ1; Z2Þ ¼ Za1 Z
12a2 ; where Zj denotes consumption of
good j: In Appendix C, we show that with this specification for utility, national
welfare in the initial steady state is given by WSS ¼ hðgÞYSS=r where
hðgÞ ¼aa½ð12 aÞð1þ gÞ�12a
aþ ð12 aÞð1þ gÞ
3.2.3. Adjustment
Changes in world prices cause the VjT curves in Figure 3.3 to pivot with the point
at which VjT ¼ 0 remaining fixed. Thus, if sector 1 is initially protected by a tariff,
then when trade is liberalized V1T pivots down causing aH to fall. If we use aFT to
denote the new value of aH (with the sub-script referring to “free trade”), then all
workers with ability levels in the interval ½aFT; aH� eventually want to switch fromthe low-tech to the high-tech sector. Trainees switch immediately while those
employed in the low-tech sector must decide whether to quit their jobs and switch
sectors immediately or keep their jobs and switch only after losing their jobs. If we
use aQ to denote the ability level of the low-tech worker who is just indifferent
between quitting and keeping his/her job, then it is straightforward to show that
aQ [ ½aFT; aH�: Employed workers with a [ ½aQ; aH� quit immediately and start
to train for high-tech jobs while those with a [ ½aFT; aQ� wait and switch after
losing their low-tech job.
C. Davidson and S.J. Matusz44
Because of the model’s simple structure, it is possible to solve analytically for
the transition path between steady states. We begin by noting that all V terms
jump immediately to their new steady state values once trade is liberalized. This
is due to the fact that these values depend only on prices, ability, turnover rates
and other parameters that are independent of time (see Equations 6–10). Thus,
aH jumps to its new value immediately as well. The gradual transition to the
new steady state occurs in the labor market and involves only those workers
with ability levels in the range ½aFT; aH�: For these workers, the measures of
trainees, searchers and employed workers change according to the differential
equations in Equations 1–5. We provide the solution to this system of differ-
ential equations in Appendix A and show how they can be used to calculate the
net output produced by these workers in each sector during the adjustment
process. We use XjðaFT; aQ; tÞ to denote the value of the net output produced in
sector j by workers with a [ ½aFT; aQ� at time t and use XjðaQ; aHÞ to play the
same role for those workers with a [ ½aQ; aH�: These values are given by
Equations A11, A12 and A15 in Appendix A.
Liberalization does not alter the behavior of those workers with a # aFT or
a $ aH: The former group remains attached to the low-tech sector whereas the
latter group continues to train, search and work in the high-tech sector. It is
possible that as workers flow from the low-tech sector to the high-tech sector,
labor market congestion might cause the job acquisition rate in the high-tech
sector to fall. If congestion externalities are present, they would increase the cost
of adjustment and slow down the transition to the new steady state. However, as
we mentioned in Section 3.1, in order to keep our analysis as simple and tractable
as possible, we abstract from this issue by assuming that all workers seeking
high-tech jobs continue to find them at the steady state job acquisition rate of e:This assumption ensures that our estimate of adjustment costs will be on the
conservative side. It also implies that for workers with a # aFT the measures of
workers training and employed in the low-tech sector after liberalization are given
by Equations 14 and 15, respectively, with ðaH 2 aLÞ replaced by ðaFT 2 aLÞ:We
use LFT1T and LFT1E to represent these values. Note that for those workers with
a $ aH; the measures training, searching and employed in the high-tech sector
after liberalization are still given by Equations 16 and 17.
Given the solutions provided in Appendix A, we can calculate the value of
output net of training costs along the transition path, YðtÞ: We have
YðtÞ ¼ p1{0:5q1ðaL þ aFTÞLFT1E 2 c1L
FT1T}þ p2{0:5q2ð1þ aHÞL2E 2 c2L2T}
þ X1ðaFT; aQ; tÞ þ X2ðaFT; aQ; tÞ þ X2ðaQ; aH; tÞ ð21Þ
Should Policy Makers be Concerned About Adjustment Costs? 45
To find welfare after liberalization, taking the adjustment path into account, we
first transform net output into utility and then integrate over time. Given our
specification of utility, we obtain WA ¼Ð
e2rthð0ÞYðtÞdt:The last step in solving for the cost of adjustment is to compare WA with the
welfare that the economy could achieve if it were able to jump immediately to the
new (free trade) steady-state equilibrium ðWFTÞ: To find this value, define YFT as
the value of output net of training costs in the new steady-state equilibrium. This
value is given by Equation 20 with aFT replacing aH so that WFT ¼ hð0ÞYFT=r:10
Typical time paths for YSS;YðtÞ and YFT are depicted in Figure 3.4. Liberalizing
trade increases steady state net output from its initial value of YSS to its new free
trade value of YFT: However, to reach the new steady state, the economy must first
go through a costly transition with net output following along the YðtÞ path. Note
that there is a period of time (up to tp) during which net output falls below its
initial steady state value. The potential gain from trade reform is defined by the
properly discounted area below YFT and above YSS; or, WFT 2WSS: The actual
gain is the properly discounted difference between the areas below YSS and YðtÞ;
or, WA 2WSS: It follows that aggregate adjustment costs are measured by the
appropriately discounted area below YFT but above YðtÞ; or, WFT 2WA:11 In
Section 3.2.4, we simulate the model, calculate aggregate adjustment costs and
compare them to the potential gains from reform.12 We do so by focusing on two
key variables. The first variable is tp;which measures the length of time it takes for
Figure 3.4: The value of output net of training costs over time.
10 Note that we are using the compensating variation to measure the change in welfare due to
liberalization.11 This method for calculating adjustment costs was suggested by Neary (1982).12 In Davidson and Matusz (2001), we explore other aspects of the adjustment path including the time
paths of employment and unemployment during the adjustment period.
C. Davidson and S.J. Matusz46
the economy to get back to its original level of net income (so that tp solves
YSS 2 YðtpÞ ¼ 0). By looking at tp we are able to get some sense as to how long it
takes the economy to begin to reap the benefits from liberalization. The second
variable of interest is Rp; defined as the ratio of aggregate adjustment costs to the
potential benefits from trade reform:
Rp;
WFT 2WA
WFT 2WSS
ð22Þ
3.2.4. Strengths and weaknesses of our model
At this point, it is useful to highlight some features of our model that we consider
strengths as well as some of the weaknesses. There are several attractive features
that are worth emphasizing. First, we have modeled the training and search
processes that workers must go through in order to find jobs. This allows us to take
into account both the time and resource costs that dislocated workers must incur
after losing their jobs. This is a unique and innovative feature of our model and we
consider it one of its main strengths. The second important feature is that we have
managed to keep the framework relatively simple and tractable. In fact, it is so
simple that we can solve explicitly for the transition path between steady states by
solving the differential equations in Equations 1–5.
Another attractive feature of our model is that many of the key parameters, for
example, the labor market turnover rates, are observable. This makes it easy to
calibrate the model and find estimates of aggregate adjustment costs for parameter
values that have some empirical significance. Moreover, as we emphasized in
Section 3.1, it is well known that labor markets in Europe, the United States and
Japan differ significantly in their structure and that much of the difference has to
do with differences in turnover rates. Since it is these turnover rates that drive our
analysis, we can easily model the differences in labor market structure across these
regions and see how our estimate of adjustment costs relative to the benefits from
trade liberalization vary with labor market flexibility.
Finally, there is one other positive feature of our model that we would like to
underscore. In Appendix B we show that the equilibrium in our model is efficient.
This is unusual for search models. It is usually the case that search decisions are
rife with externalities. For example, if an unemployed worker chooses to seek a job
in a particular sector, this maymake it more difficult for other unemployed workers
to find a job (i.e., there may be congestion externalities). Such externalities
typically distort behavior and lead to sub-optimal equilibria. This is not the case in
our model. In fact, we set up our model with certain features (such as exogenous
Should Policy Makers be Concerned About Adjustment Costs? 47
job acquisition rates) specifically designed to avoid this problem. The reason that
we did this is so that when we calculate adjustment costs and compare them to the
gains from trade we can be certain that our results do not depend on how trade
liberalization affects the distortions created by controversial, hard-to-measure
search generated externalities.
While there is considerable merit in assuming fixed job acquisition rates, there
is also a downside. Agents face changing economic incentives as trade is
liberalized. Some workers in the import competing sector quit and this results in
an immediate increase in the job destruction rate. At the same time, since the pool
of unemployed workers is swelling in the export sector, the job acquisition
rate may fall due to congestion. As mentioned in Section 3.1, this possibility has
been the focus of several papers in which it has been suggested that gradual
liberalization may be warranted in order to avoid congestion in export sector
labor markets. By treating the job acquisition rate as an exogenously specified
parameter, we are ignoring these possibilities and obtaining estimates of
adjustment costs that are probably too low.
Another weakness in our analysis concerns the parameters that measure the
costs of retraining (cj and tj). Although these parameters play a key role in our
analysis, we know very little about their likely values. We handle this problem in
two ways. First, since it is unlikely that training costs in the low-tech sector are
significant, we set c1 equal to zero and assume that t1 is quite high (so that low-
tech training is very brief). Given that we have also assumed that there are no
resource costs associated with job search, this is another reason to suspect that our
estimates of aggregate adjustment costs are likely to be biased downward. Second,
we consider a wide variety of assumptions about the magnitude of high-tech
training costs and then try to draw conclusions that are robust across these sets of
assumptions.
3.3. AGGREGATE ADJUSTMENT COSTS
The parameters of our model include those that determine the average durations of
sector-j training ðtjÞ; sector-j employment ðbjÞ and high-tech search ðeÞ; those thatdetermine the resource cost of sector-j training ðcjÞ; those that help to determine
output per worker in sector j ðqjÞ; the rate of transferability of high-tech skills
across jobs ðfÞ; the discount rate ðrÞ; and the share of income devoted to
consumption of good 1 ðaÞ: In this section, we choose values for these parameters
to solve the model and provide measures of the aggregate adjustment costs
associated with trade reform. We do so under the assumption that the low-tech
sector is initially protected by a 5% tariff (thus, g ¼ 0:05).
C. Davidson and S.J. Matusz48
To make certain that we do not discount the future too heavily, we set r ¼ 0:03;the lowest discount rate considered by Baldwin et al. (1980) and Magee (1972).13
The average duration of unemployment in the US can be found in The 2001
Economic Report of the President (see Table B-44). Although this value has
fluctuated over the years, it remains fairly stable at about one quarter (or 13 weeks).
Our model is consistent with such estimates if we set e ¼ 4: Since this value rarelyfluctuates by more than a week or two, this is the only value for e that we consider.
For the average duration of employment in the high-tech sector, we turn to the job
creationanddestructiondataofDavis et al. (1996),who report that the average annual
rate of job destruction in USmanufacturing during the period 1973–1988 was about
10% (this translates into an average job duration of 10 years). There is some variation
in this number across years, with the largest rate of job destruction coming in 1975 at
16.5% (implying an average job duration of about 6 years).14 We therefore assume,
for our base case, that an average high-tech job lasts 10 years (which is the case if
b2 ¼ 0:10).However,we also solve themodel and report results for the case inwhich
high-tech jobs last only 6 years (which is the case if b2 ¼ 0:167).It is hard to find data on the average duration of a job in the low-tech sector. We
consider these to be transitory, undesirable jobs and although many of these jobs
may be found in the manufacturing sector, it is not possible to look at industry-
wide data and draw conclusions about how long the worst jobs in each industry
last. So, we take a different approach. Our low-tech jobs require few skills and
little training. These are the types of jobs that many hold while still in school or
when they are just starting out in the labor force. If we look at data on the number
of jobs held over the lifetime, we find that up to the age of 24 workers hold
(roughly) one new job every 2 years.15 We therefore consider two cases—one in
which low-tech jobs last 2 years (so that b1 ¼ 0:5) and another in which they last
just 1 year (so that b1 ¼ 1:0). Combining these two cases with the assumptions
that we have made about job tenure in the high-tech sector leaves us with four
different settings. In the setting with high turnover in both sectors, jobs last a year
in the low-tech sector and 6 years in the high-tech sector. In the setting with low
turnover in both sectors, jobs last 2 years in the low-tech sector and 10 years in the
high-tech sector. In the other two cases, jobs last either 3 or 10 times as long in the
high-tech sector than they do in the low-tech sector. This gives us a wide range of
assumptions about labor market turnover.
13 This again biases our result in terms of minimizing the magnitude of adjustment costs, since a
low discount rate places relatively little weight on current costs versus the future benefits of
reform.14 See Table 2.1 on p. 19 in Davis et al. (1996).15 See Table 8.1 on p. 210 in Hamermesh and Rees (1988).
Should Policy Makers be Concerned About Adjustment Costs? 49
Turn next to the parameters of the training processes. Since very little is known
about themagnitude of training costswewant to be careful not to assume values that
seem unreasonably high, andwewant tomake sure that we consider awide range of
possible values. Aswementioned above, we assume that there are no resource costs
associated with low-tech training (i.e., c1 ¼ 0).16 In addition, we assume that the
low-tech training process takes only 1 week (so that t1 ¼ 52). For the high-tech
sector,we turn to the limited information that is available on training costs.A review
of what is known about turnover costs can be found in Hamermesh (1993) where
turnover costs are assumed to include both the costs of recruiting and training the
newly hired worker. This literature suggests that such costs may be quite high. For
example, a large firm in the pharmaceutical industry estimated that the present
value of the cost of replacing one worker amounted to roughly twice that worker’s
annual salary. Similar, although not quite so dramatic, estimates were obtained for
less-skilled jobs. One study estimated that the cost of replacing a truck driver
amounted to slightly less than half of that worker’s annual pay. The lowest estimate
of turnover costs reported by Hamermesh appears to be about 3 weeks worth of
salary. Similar estimates can be found in Acemoglu and Pischke’s (1999) study
of the training process in the German apprentice system. They report estimates of
training costs that vary from 6 to 15months of the average worker’s annual income,
depending on the size of the firm. To capture this wide range of estimates, we
assume that high-tech training lasts 4months ðt2 ¼ 3Þ and thenwe vary the value of
c2:At the low end, we choose c2 so that training costs for the average worker in the
high-tech sector are equal to 1 month’s pay.17At the high end, we choose c2 so that
the average high-tech worker’s training costs equal 15 months pay. We also
consider two intermediate cases inwhich training costs equal 5 and 10months of the
worker’s annual salary. This gives us a wide range of values for high-tech training
costs. Below we look for results that are robust across this range of estimates.18
16 With c1 ¼ 0 we have aL ¼ 0 so that all workers enter the labor market.17 High-tech workers pay a flow cost of p2c2 while training and training lasts, on average, 1=t2 periods.
Thus, training costs are given by p2c2=t2:Annual income for the average worker in the high-tech sector
is p2q2ðaH þ aLÞ=2:18 At this point, it is useful to first clarify what we mean by training costs. While acquiring the skills
necessary to perform certain tasks, there may be periods during which no production occurs whatsoever
(while workers are in school, going through orientation, getting hands-on on-the-job training, and
so on). However, there may also be a period during which the worker is producing and yet productivity
is below its ultimate level because the worker is still learning about the production process. The output
lost during the period of learning-by-doing should also be considered as part of training costs. With this
interpretation, it is hard to imagine that our most modest assumptions—that there are no resource costs
to training in the low-tech sector, that the low-tech training process takes only 1 week, and that high-
tech training costs amount to only 1 month’s worth of high-tech wages—could be considered
excessive.
C. Davidson and S.J. Matusz50
This leaves q1 and q2; the productivity parameters in the two sectors, f; whichmeasures how often high-tech workers need to retrain after losing their jobs, and
a; the parameter in the utility function. Our simulations indicate that our results
are quite insensitive with respect to a. Adjustment costs are minimized for a ¼
1=2 and rarely vary by more than 0.02 for other values of a (relative to the values
in Table 3.1). As for f; we have argued that high-tech jobs require both general
and job-specific training with much of the training general. The implication is that
retraining is not all that common in the high-tech sector, which means that fshould be fairly high. In Tables 3.1 and 3.2 we provide estimates of the two
variables that we are interested in, Rp and tp; under the assumption that f ¼ 0:8:However, we also calculated these values assuming that f ranged between 0.5 and
0.9 and found that the values in Tables 3.1 and 3.2 were affected only at the third
decimal place. Thus, we conclude that our estimates are also largely insensitive to
our assumptions about f; provided that this value remains above 0.5.
For q1 and q2 what matters is their relative value. Thus, we set q2 ¼ 1:4 and thenvary q1:As q1 varies, the relative attractiveness of the two sectors changes and thus,
Table 3.1: Aggregate adjustment costs as a fraction of the gross benefits from trade reform ðRpÞ:
Training costs aH
0.20 0.33 0.50 0.66 tp
b1 ¼ 0:5; b2 ¼ 0:11 month 0.42 0.39 0.36 0.34 1.3–1.4
5 months 0.66 0.63 0.57 0.53 1.5–1.8
10 months 0.75 0.74 0.69 0.65 1.7–2.1
15 months 0.78 0.80 0.76 0.72 1.9–2.3
b1 ¼ 0:5; b2 ¼ 0:1671 month 0.42 0.39 0.36 0.34 1.3–1.4
5 months 0.65 0.63 0.58 0.53 1.5–1.9
10 months 0.73 0.74 0.70 0.65 1.8–2.2
15 months 0.76 0.80 0.76 0.72 2.0–2.3
b1 ¼ 1; b2 ¼ 0:11 month 0.42 0.39 0.36 0.34 1.4–1.5
5 months 0.66 0.63 0.57 0.53 1.8–2.3
10 months 0.75 0.74 0.69 0.65 2.1–2.6
15 months 0.78 0.80 0.76 0.72 2.4–2.7
b1 ¼ 1; b2 ¼ 0:1671 month 0.42 0.39 0.36 0.34 1.4–1.6
5 months 0.65 0.63 0.57 0.53 1.8–2.3
10 months 0.73 0.74 0.70 0.65 2.2–2.6
15 months 0.76 0.80 0.76 0.72 2.4–2.7
Should Policy Makers be Concerned About Adjustment Costs? 51
aH; which determines the fraction of the workforce that starts out in the low-tech
sector, is altered. For completeness, we consider four different values for q1 for
each combination of turnover rates. These are the values that correspond to aHequal 0.2, 0.33, 0.5, and 0.66. This gives us a sense as to how our measures of Rp
and tp vary with the size of the sector that is initially protected (sector 1) relative to
the size of the sector that is associated with significant training costs (sector 2).
Our estimates of Rp and tp; are reported in Table 3.1. They were obtained by
assuming that the world prices of the two goods are the same and that the low-tech
sector is initially protected by a 5% tariff. Three results stand out. First, our model
predicts that adjustment will take place quickly, with output reaching its pre-
liberalization level in about 2 years. The immediate implication is that if one were
to look for evidence of adjustment costs using yearly data, it would appear that
such costs are quite low.
Nevertheless, the second result that stands out is that our estimates of
adjustment costs are considerably higher than any obtained by Baldwin et al.
(1980) or Magee (1972). Our lowest estimate in Table 3.1 is that adjustment
Table 3.2: Aggregate adjustment costs as a fraction of the gross benefits from trade reform ignoring
the resource costs from high-tech training ðRp
GOÞ:
Training costs aH
0.20 0.33 0.50 0.66 tpGO
b1 ¼ 0:5; b2 ¼ 0:11 month 0.20 0.23 0.24 0.25 1.2–1.3
5 months 0.12 0.16 0.18 0.20 1.0–1.2
10 months 0.08 0.11 0.14 0.16 0.8–1.1
15 months 0.06 0.09 0.11 0.13 0.7–1.1
b1 ¼ 0:5; b2 ¼ 0:1671 month 0.19 0.22 0.23 0.24 1.2
5 months 0.10 0.14 0.17 0.18 0.9–1.2
10 months 0.07 0.10 0.12 0.14 0.7–1.1
15 months 0.05 0.08 0.10 0.12 0.6–1.0
b1 ¼ 1; b2 ¼ 0:11 month 0.20 0.23 0.24 0.25 1.2–1.3
5 months 0.11 0.16 0.18 0.20 1.1–1.3
10 months 0.08 0.11 0.14 0.16 0.9–1.2
15 months 0.06 0.09 0.12 0.13 0.8–1.2
b1 ¼ 1; b2 ¼ 0:1671 month 0.19 0.22 0.23 0.24 1.2–1.3
5 months 0.10 0.14 0.17 0.18 1.0–1.3
10 months 0.07 0.10 0.12 0.14 0.8–1.2
15 months 0.05 0.08 0.10 0.12 0.6–1.1
C. Davidson and S.J. Matusz52
costs eat away about one third of the gains from trade reform. At the other
extreme, some estimates are as high as 80%! Given that we have assumed
away search costs and resource costs for low-tech training and abstracted from
congestion externalities in the expanding export sector labor markets, these
estimates are surprisingly high.
Third, the results with respect to Rp and tp are remarkably robust across our
assumptions about steady state break-up rates—going from high turnover in both
sectors to low turnover in both sectors never changes Rp by more than 0.02. Our
estimates are also fairly insensitive to our assumptions about the initial size of the
low-tech sector. As aH increases, Rp and tp tend to fall with the rate of decrease
increasing in the magnitude of high-tech training costs.19 In fact, it is the
magnitude of these training costs that clearly matter the most. Not surprisingly, as
training costs increase, so do Rp and tp:One natural question to ask at this point is whether our results are driven by our
assumption that training involves a real resource cost or whether the costs are this
high simply because the training and search processes take time and no production
occurs while search and training take place. To get some handle on this issue,
we introduce two new terms, Rp
GO and tpGO: These terms are defined in exactly the
same manner as Rp and tp with one exception—they measure only gross output
(i.e., they ignore the resource cost of training). So, for example, tpGO measures the
amount of time it takes for gross output to get back to its pre-liberalization level.
Our estimates of Rp
GO and tpGO are reported in Table 3.2. While our estimates fall
significantly, they remain considerably above those found in previous studies.
Most of the estimates indicate that when we take into account only the time cost of
training, around 15–20% of the gains in gross output are lost due to adjustment
costs. Moreover, gross output returns to its pre-liberalization level before net
output, making it even harder to find evidence of significant adjustment costs in
annual (or even quarterly) data. These estimates are robust across our assumptions
concerning break-up rates and the initial size of the low-tech sector, but do vary
significantly as we change our assumptions about the magnitude of high-tech
training costs.
3.4. ADJUSTMENT COSTS AND LABOR MARKET FLEXIBILITY
Tables 3.1 and 3.2 indicate that changes in break-up rates have little influence over
our estimates of aggregate adjustment costs. Yet, if we look across the world, it is
19 Increasing the initial size of the low-tech sector has two effects on Rp: On one hand, if the low-tech
sector is large then trade reform will generate large benefits. On the other hand, a large low-tech sector
trade reform will also lead to a great deal of worker reallocation and this will increase adjustment costs.
Should Policy Makers be Concerned About Adjustment Costs? 53
not only break-up rates that vary but also the rate of job acquisition. In the US
most unemployed workers find reemployment relatively quickly and long-term
unemployment is not a significant problem. In contrast, many European
economies face serious problems with a large population of workers who have
been classified as long-term unemployed. Combining this with the fact that job
duration is also longer in Europe leads to the conclusion that labor markets are
much more flexible in US than they are in Europe. This difference in labor market
flexibility has been emphasized by labor economists and macroeconomists
studying a variety of issues.20 In this section, we investigate the implications for
aggregate adjustment costs.
To do so, we add a new variable s to our model, which we refer to as speed. We
introduce this term by multiplying the turnover rates in the high-tech sector, b2and e; by s:21 As s increases, high-tech jobs become easier to find but they also
become less durable. Thus, an economy with a high value for s has a great deal of
turnover in the high-tech sector while an economy with a low value for s has a
high-tech sector with a long average duration of unemployment and a relatively
long expected job tenure. It follows that s measures the flexibility of the labor
market with increases in speed associated with more flexible labor markets.22
Figure 3.5a shows how Rp varies with s for the case in which there is low
turnover in the high-tech sector, high turnover in the low-tech sector, high-tech
training costs are equal to 5 months of the average high-tech worker’s income and
one-third of the labor force starts out in sector 1 (i.e., b1 ¼ 1:0; b2 ¼ 0:1 and
aH ¼ 0:33). Qualitatively similar figures apply for all other parameter values in
Tables 3.1 and 3.2. As expected, there is a negative relationship between the two
measures—increases in labor market flexibility always reduce relative adjustment
costs. The implication is that Americans should be less concerned about the costs
of adjustment than Europeans and/or the Japanese.
The negative relationship depicted in Figure 3.5a suggests that it would be
useful to look at how the actual gains from trade (net of adjustment costs) vary
with s: To do so, define NB to be the benefit from liberalization net of adjustment
20 See, for example, Freeman (1994) and Layard et al. (1991).21 Similar results can be obtained by multiplying all turnover rates by s so that an increase in s results in
higher turnover in both sectors. However, since turnover plays a more prominent role in the high-tech
sector, it turns out that changes triggered by changes in the high-tech rates dominate those driven by
changes in the low-tech rates. A brief description of this case can be found in Davidson and Matusz
(2000).22 Note that we do not multiply the turnover rates associated with training by s: It is our view that the
length of the training process is determined by the complexity of the job and this is a feature that is
linked to technology, not the flexibility of the labor market.
C. Davidson and S.J. Matusz54
costs (as a percentage of initial steady state welfare):
NB ; 100WA 2WSS
WSS
� �
ð23Þ
Figure 3.5b shows how NB varies with s for the parameter values that generate
Figure 3.5a. Surprisingly, as with Rp; the relationship is negative. Even though
economies with the least flexible labor markets have the highest adjustment costs,
Figure 3.5: (a) Rp as a function of s: (b) NB as a function of s:
Should Policy Makers be Concerned About Adjustment Costs? 55
as indicated by high values of Rp; they have the most to gain from liberalization, as
indicated by high values of NB.
The surprising outcome depicted in Figure 3.5b can be traced to the manner in
which the gross benefit from trade varies with speed. The gross benefit from trade
reform depends on the amount of workers who switch sectors as a result of
liberalization. The more workers that switch, the greater the increase in gross
output. Figure 3.3 can be used to see how the amount of worker reallocation varies
with s: Trade reform lowers the return to training in the low-tech sector, causing
the V1T curve to pivot down. The amount of worker reallocation that occurs then
depends on the slope of the V2T curve with a flatter curve implying more
reallocation. It is straightforward to show that this curve is relatively flat when s is
either very low or very high. Intuitively, if s ¼ 0 a worker who is currently
training has no hope of ever finding a job (since the job acquisition rate is 0) and
thus ability, which only affects output while employed, plays no role in
determining the value from training. In this case, the V2T curve is horizontal.
Increasing s leads to an increase in the fraction of life spent employed and makes
ability more important. Thus, when s is low, an increase in s causes the V2T curve
to become steeper. It follows that trade reform results in a great deal of
reallocation when s is very low—thus, when the tariff is removed, large benefits
accrue to economies with slothful labor markets. In fact, Figure 3.5b reveals that
the benefits are so large that they swamp the high costs of adjustments that these
economies face.
These results can be viewed one of two ways. On one hand, there is good
news for economies with slothful labor markets—they have much to gain
from trade reform, even though they will face high costs of adjustment during
the transition to the new steady state. However, the reason that they gain so
much is the bad news—in such economies, tariffs have large distortionary
effects because they cause a great deal of worker reallocation. Removing the
tariff therefore generates gross benefits that are large enough to swamp the
adjustment costs.
It does, however, take time for the economy with the slothful labor markets
to realize these large gains and since turnover is low it may be quite some
time before net output returns to its pre-reform level. For example, for the
case reported in Table 3.1, while it takes 2 years for the base case economy
ðs ¼ 1Þ to get net output back to its initial level, it takes an additional 2.5
years for the most stagnant economy (with s ¼ 0:15 we find that tp ¼ 4:5).The implication is that although the gains from liberalization may be quite
large in such economies, it may be very difficult to find any politician willing
to push for such reform.
C. Davidson and S.J. Matusz56
Recent evidence suggests that one possible way to interpret these results would
be to have the US play the role of the flexible economy, Western European
countries (e.g., France, Belgium and the UK) play the role of the sluggish
economies and countries in Eastern Europe (e.g., Estonia, Slovenia, Bulgaria,
Hungary and Romania) play the role of the slothful economies.23 The case of
Estonia is particularly noteworthy. Haltiwanger and Vodopivec (2000) provide
evidence that at the time of significant price and trade reforms (in 1989) Estonian
labor markets were essentially stagnant. Shortly after instituting these reforms, the
Estonian government also began to implement policies aimed at increasing the
flexibility of their factor markets. As a result, the economy suffered huge costs in
the short run, with real output falling a cumulative 31% between 1991 and 1993.
However, the massive reallocation seemed to be nearing completion by 1994 and
real output has risen steadily since then. Estonia is largely viewed as a major
success story and our results provide some insight as to why they have been so
successful. The initial stagnant nature of their factor markets indicated that they
had a tremendous amount to gain from reform. In addition, by increasing the
flexibility of their labor markets they have been able to realize these gains much
quicker than other transition economies.
3.5. CONCLUSION
There is no dispute about the fact that workers lose jobs due to changes in trade
patterns and that protecting an industry saves jobs. For example, Hufbauer and
Elliott (1994) estimate that eliminating protection in the US apparel industry
would cost over 150,000 workers their jobs. It is also well documented that
dislocated workers suffer large personal losses with some estimates for the
average loss as high as $80,000 in lifetime earnings (Jacobson et al., 1993a,b).
It is therefore not surprising that political leaders are sometimes hesitant about
trade reform. Those who lose may lose a great deal and are likely to remember
who is at fault when the next election nears. The gains are delayed and are
spread out over many so that those who do gain probably gain much less than
the few who lose.
Nevertheless, there is probably no other position in economics that has as much
widespread support as the belief in the benefits from freer trade. Academic
economists typically respond to public concerns about the losses to dislocated
workers by explaining that such concerns are misplaced and misguided. This view
23 See Haltiwanger and Vodopivec (2000) for a discussion of job flows in Estonia and Slovenia and
Bilsen and Konings (1998) for a discussion of job flows in Bulgaria, Hungary and Romania.
Should Policy Makers be Concerned About Adjustment Costs? 57
was summarized and, we feel, appropriately criticized by Baldwin et al. (1980) in
their article on adjustment costs
Economists have sometimes dismissed such adjustment costs with the
comment that the displaced factors become reemployed ‘in the long run’.
But this is bad economics, since in discounting streams of costs and
benefits for welfare calculations, the near-present counts more heavily
than ‘the long-run’.
In this chapter, we have tried to take a serious look at the possible magnitude of
the adjustment costs that are likely to arise from trade reform. The novelty of our
approach is that we have modeled the training and search processes that workers
must go through in order to find jobs. This allows us to take into account the time
and resource costs of retraining and job search. We have tried to be modest in our
assumptions concerning these costs. We have assumed away the resource costs
associated with job search and low-tech training and assumed that the time costs
involved in low-tech training are very small. We have also assumed that workers
fleeing the import-competing sector do not encounter congestion when trying to
secure reemployment in the export sector. Finally, we have looked at a wide
variety of assumptions concerning the cost of training in the high-tech sector.
Our results are surprising. Even with our most modest assumption concerning
the cost of high-tech training (that they equal 1 month of the average high-tech
worker’s annual earnings), we find that adjustment costs are a significant fraction
of the gross benefits from trade reform. Our lowest estimate is that roughly 30% of
the gross benefits will be eaten away by adjustment. At the other extreme, we find
that when high-tech training is costly (15 months of the average worker’s annual
salary) as much as 80% of the gross benefits may disappear during the transition
period. Even when we focus attention on gross output (so all that matters are the
time costs of training and job search) we find that our estimates of adjustment costs
are at least twice as high as previous estimates (Magee, 1972; Baldwin et al.,
1980). We also find that the transition period may be quite short, with net output
returning to its pre-liberalization level in 2.5 years or less. We have argued that
such quick adjustment maymask the true nature of the adjustment process, perhaps
implying that adjustment costs are low when they are, in fact, quite high.
In the latter part of the chapter, we investigate the relationship between labor
market flexibility, the gains from trade reform and aggregate adjustment costs. It is
well documented that turnover rates vary significantly across countries (Freeman,
1994). Part of the reason for this is that countries vary in generosity of the social
safety nets they provide for the poor and the jobless. Firing costs and generous
unemployment insurance programs contribute to long-term unemployment and low
C. Davidson and S.J. Matusz58
turnover throughout Europe (Ljungqvist and Sargent, 1998). In addition, the
widespread influence of unions in Europe contrasts sharplywith their role in theUS,
resulting in more rigid wages in European labor markets. Labor and macro-
economists have recognized that this difference in labor market structure has
important implications for issues such as job training and macroeconomic
performance (see, for example, Layard et al., 1991). As far as we know, we are
the first to investigate the implications for the net gains from trade liberalization.
Again, our main result could not have been anticipated. We find that tariffs
create the biggest distortions in economies with slothful labor markets. As a result,
when trade is liberalized these economies have the most to gain. This is true in
spite of the fact that adjustment costs are high when there is very low turnover.
Thus, our analysis indicates that policy makers in economies with slothful labor
markets should not be reluctant to reduce barriers to trade, even though their
economies are likely to face high adjustment costs.
We close with a word of caution about the interpretation of our results.
Although we have argued that adjustment costs are probably higher than previous
studies indicate, it is still clear that trade liberalization is the correct path to take—
after all, adjustment costs, although high, are always less than the gains from
reform. However, what our results do imply is that we should take more seriously
the issue of how to compensate those who bear the burden of adjustment and those
who lose when trade barriers are removed. In addition, it would be worthwhile to
investigate the manner in which various labor market policies affect the speed
with which the economy makes the transition to free trade and the manner in
which these policies affect the distribution of income during the transition period.
Our results in Section 3.4 indicate that the answers to these questions will be
particularly important for countries with slothful labor markets.
Finally, there is one other important lesson that can be gleaned from our
analysis—the costs associated with new protectionist measures are probably
higher than previously imagined. Not only do such measures distort the economy,
but our results also imply that the cost of moving from an initial steady state to a
new one characterized by higher trade barriers is quite high. This gives yet another
reason to resist protectionist policies. Moreover, our analysis in Section 3.4
indicates that it is economies with slothful labor markets that have the most to lose
from new policies that restrict trade.
ACKNOWLEDGEMENTS
We are grateful to Spiros Bougheas, Jonathan Eaton, Rodney Falvey and David
Greenaway for their insightful comments on earlier drafts.
Should Policy Makers be Concerned About Adjustment Costs? 59
APPENDIX A
In this section we show how to solve the differential equations in Equations 1–5
and obtain a closed form solution for the transition path to the new steady-state
equilibrium. As discussed in the text, workers with ability levels in the intervals
ðaL; aFTÞ and ðaH; 1Þ do not change their behavior after liberalization. Those in theformer interval remain attached to sector 1 while those in the latter interval remain
attached to sector 2. It follows that the measure of workers in these intervals that
are training or employed are given by Equations 14–17 with the aH term in
Equations 14 and 15 replaced by aFT:The remaining workers, those with ability levels in the interval ðaFT; aHÞ; want
to switch from sector 1 to sector 2. Some will do so immediately, either because
they are training at the time of liberalization or because they choose to quit their
low-tech job, whereas others opt to keep their low-tech job and make the switch
after that job dissolves. We refer to all of these workers as the “switchers”. To
figure out howmany switchers are in each labor market state at time t;we begin byintroducing some new notation. We define S
QT ðtÞ as the measure of workers with
a [ ½aFT; aQ�who switch from sector 1 to sector 2 following liberalization and are
training in sector 2 at time t: Similarly define SQS ðtÞ as the measure of workers with
a [ ½aFT; aQ� who switch from sector 1 to sector 2 and are searching at time t and
we use SQEjðtÞ to denote the measure of switchers with a [ ½aFT; aQ� who are
employed in sector j at time t: Finally, we define SHT ðtÞ; SHS ðtÞ; and SHEjðtÞ
analogously for those switchers with a [ ½aQ; aH�:Since we are using aQ to denote the ability level of the switcher who is just
indifferent between quitting and not quitting his/her low-tech job after liberal-
ization, aQ is the value of ai that equates V2TðaiÞ with V1EðaiÞ after liberalization
(with V1Eðai) adjusted to take into account the fact that the worker switches sectors
after losing his/her low-tech job). It follows that all workers with a [ ½aQ; aH�switch sectors immediately after liberalization, whereas workers with a [
½aFT; aQ� who are employed when the tariff is removed retain their job until it
dissolves. Then, for those switchers with a [ ½aFT; aQ�; the system of differential
equations can be written as in Equations A1–A4
_SQE1
¼ 2b2SQE1
ðA1Þ
_SQE2
¼ eSQS 2 b2S
QE2
ðA2Þ
_SQS ¼ b2fS
QE2
þ t2SQT 2 eS
QS ðA3Þ
ðaQ 2 aFTÞL ¼ SQE1
þ SQE2
þ SQS þ S
QT ðA4Þ
C. Davidson and S.J. Matusz60
where for notational convenience, we have suppressed the time argument.
Equation A1 is a simple differential equation, which has the following solution
SQE1ðtÞ ¼
t1t1 þ b1
ðaQ 2 aFTÞLe2b1t ðA5Þ
In solving Equation A1, we make use of the initial condition that
SQE1ð0Þ ¼
t1t1 þ b1
ðaQ 2 aFTÞL
To solve Equations A2–A4, substitute Equation A5 into Equation A4, solve for
SQT in terms of SQE2
and SQS and then substitute the result into Equation A3. This
leaves us with Equations A2 and A3 which form a system of two differential
equations that can be written in matrix form
_SQE2
_SQS
2
4
3
5 ¼2b2 e
b2f2 t2 2ðeþ t2Þ
" #SQE2
SQS
2
4
3
5þ0
hðtÞ
" #
ðA6Þ
where
hðtÞ ; t2LðaQ 2 aFTÞ 12t1
t1 þ b1e2b1t
� �
The method for solving a system of this form can be found in Boyce and DiPrima
(1977, pp. 329–331). Using the initial conditions that SQE2ð0Þ ¼ S
QS ð0Þ ¼ 0; the
solutions are
SQE2ðtÞ ¼
et2ðaQ 2 aFTÞL
l2 2 l1
el2t
l22
el1t
l1
" #
þet1t2ðaQ 2 aFTÞL
ðt1 þ b1Þðl2 2 l1Þ
�el1t
l1 þ b12
el2t
l2 þ b1
" #
2et2ðaQ 2 aFTÞL
l2l1
2et1t2ðaQ 2 aFTÞL
ðt1 þ b1Þðl1 þ b1Þðl2 þ b1Þe2b1t ðA7Þ
Should Policy Makers be Concerned About Adjustment Costs? 61
SQS ðtÞ ¼
t2b2ðaQ 2 aFTÞL
l1l2þ
t1t2ðaQ 2 aFTÞL
ðt1 þ b1Þðl2 2 l1Þ
�b2 þ l1b1 þ l1
el1t 2
b2 þ l2b1 þ l2
el2t
� �
2t2ðaQ 2 aFTÞL
ðl2 2 l1Þ
�b2 þ l1
l1el1t 2
b2 þ l2l2
el2t
� �
2t1t2ðb2 2 b1ÞðaQ 2 aFTÞL
ðt1 þ b1Þðl1 þ b1Þðl2 þ b1Þ
� e2b1t ðA8Þ
where l1 and l2; the eigenvalues of the coefficient matrix in Equation A6, are
given by
l1 ¼2ðb2 þ eþ t2Þ2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðb2 þ eþ t2Þ2 2 4b2eð12 fÞ
p
2ðA9aÞ
l1 ¼2ðb2 þ eþ t2Þ þ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðb2 þ eþ t2Þ2 2 4b2eð12 fÞ
p
2ðA9bÞ
Finally, the measure of workers with a [ ½aFT; aQ� training in sector 2 at time t is
given by
SQT ðtÞ ¼ ðaQ 2 aFTÞL2 S
QE1ðtÞ2 S
QE2ðtÞ2 S
QS ðtÞ ðA10Þ
The value of the net output produced in each sector by this group of workers at
time t is therefore given by
X1ðaFT; aQ; tÞ ¼ 0:5ðaFT þ aQÞp1q1SQE1ðtÞ ðA11Þ
X2ðaFT; aQ; tÞ ¼ p2{0:5ðaFT þ aQÞq2SQE2ðtÞ2 c2S
QT ðtÞ} ðA12Þ
For those switchers with a [ ½aQ; aH�; the system of differentials is given by
Equations A2–A4 with the “Q” superscript replaced by “H”. Equation A1 is no
longer valid, since all of these switchers quit their jobs. Following the solution
C. Davidson and S.J. Matusz62
method described above yields
SHE2ðtÞ ¼ ðaH 2 aQÞLet2
1
l1l2þ
1
l2 2 l1
el2t
l22
el1t
l1
" #( )
ðA13Þ
SHS ðtÞ ¼ ðaH 2 aQÞLt2
b2
l1l2þ
1
l2 2 l1
b2 þ l2l2
el2t 2
b2 þ l1l1
el1t
� ��
ðA14Þ
and SHT ðtÞ ¼ ðaH 2 aFTÞL2 SHE2ðtÞ2 SHS ðtÞ: The value of the net output produced
in sector 2 by this group of workers at time t is therefore given by
X2ðaQ; aH; tÞ ¼ p2{0:5ðaH þ aQÞq2SHE2ðtÞ2 c2S
HT ðtÞ} ðA15Þ
APPENDIX B
In this section we show that the laissez-faire equilibrium in our model is efficient.
To do so, we calculate the dynamic marginal product of labor in each sector and
show that these values are equal in the market equilibrium.
The dynamic marginal product of sector j labor measures the increase in net
output that occurs if the steady state is disturbed by adding an additional worker to
that sector taking into account the adjustment path to the new steady state. To
calculate the dynamic marginal products we follow the method developed by
Diamond (1980).
We begin by defining xiðuÞ as the present discounted value of output net of
training costs produced in sector i when a (small) measure u of new workers is
added to that sector. These workers are assumed to have ability level aH:Equilibrium is efficient if x 0
1ðuÞ ¼ x 02ðuÞ:
Start with sector 1. We have24
x1ðuÞ ;ð1
0e2rt{aHq1p1uIðtÞ2 up1c1½12 IðtÞ�}dt
where _uE1 ¼ t1u2 ðt1 þ b1ÞuE1 and IðtÞ is an indicator function that takes on the
value of 1 when the worker is employed and equals zero at all other times. To find
24 The equation of motion for _uE1 is obtained in the following manner. Since search is not required to
find employment in sector 1, we have _uE1 ¼ t1uT1 2 b1u
E1 : Now, we know that the total measure of
trainers (out of the u) in sector 1 is equal to the difference between u and the measure of employed
workers in that sector. Substituting for uT1 yields the desired result.
Should Policy Makers be Concerned About Adjustment Costs? 63
x 01ðuÞ we start by using the fundamental equation of dynamic programming which
states that
rx1ðuÞ ¼ aHq1p1uIðtÞ2 up1c1½12 IðtÞ� þ›x1›uE1
_uE1
Substituting for _uE1 from above allows us to write this as
rx1ðuÞ ¼ aHq1p1uIðtÞ2 up1c1½12 IðtÞ� þ›x1›uE1
{t1u2 ðt1 þ b1ÞuE1} ðB1Þ
Differentiating with respect to u yields
rx01ðuÞ ¼ aHq1p1IðtÞ2 p1c1{12 IðtÞ}þ t1›x1›uE1
but, at t ¼ 0; IðaH; 0Þ ¼ 0 so that we have
rx01ðuÞ ¼ 2p1c1 þ t1›x1›uE1
ðB2Þ
To complete our derivation, we must now calculate ›x1=›uE1 : To do so, we
solve Equation B1 for ›x1=›uE1 : We obtain
›x1›uE1
¼rx1 2 aHq1p1uIðtÞ þ up1c1½12 IðtÞ�
t1u2 ðt1 þ b1ÞuE1
In the initial steady state, the right-hand side of this equation equals 0/0. Applying
L’Hopital’s Rule, we have (note that we are differentiating with respect to uE1 ;which is same as uIðtÞ)
›x1›uE1
¼
r›x1›uE1
2 aHq1p1 2 p1c1
2 ðt1 þ b1Þ
or
›x1›uE1
¼aHq1p1 þ p1c1
r þ t1 þ b1
C. Davidson and S.J. Matusz64
We can now substitute this value into Equation B2 to obtain the dynamic marginal
product of labor in sector 1:
rx01ðuÞ ¼t1aHq1p1 2 ðr þ b1Þp1c1
r þ t1 þ b1ðB3Þ
Note that this dynamic marginal product equals rV1TðaHÞ:We now turn next to sector 2. We have
x2ðuÞ ;ð1
0e2rt{aHp2q2uIðtÞ2 c2p2u½12 IðtÞ2 HðtÞ�}dt
where _uE2 ¼ euS2 2 b2uE2 ; _uS2 ¼ t2uþ ðb2f2 t2Þu
E2 2 ðt2 þ eÞuS2 : IðtÞ is an
indicator function that equals one when the worker is employed and zero
otherwise and HðtÞ is an indicator function which equals one when the worker is
searching and zero otherwise.
As above, we start by applying the fundamental equation of dynamic
programming which implies that
rx2ðuÞ ¼ aHp2q2uIðtÞ2 c2p2u½12 IðtÞ2 HðtÞ� þ›x2›uE2
_uE2 þ›x2›uS2
_uS2
If we now use the equations of motion to substitute for _uE2 and _uS2 and then
differentiate with respect to u we obtain
rx02ðuÞ ¼ aHp2q2IðtÞ2 c2p2½12 IðtÞ2 HðtÞ� þ›x2›uS2
t2
But, in the initial steady state (at t ¼ 0), we know that Ið0Þ ¼ Hð0Þ ¼ 0; so that
rx02ðuÞ ¼ 2c2p2 þ t2›x2›uS2
ðB4Þ
The final step requires us to solve for ›x2=›uS2 and then substitute that value into
Equation B4. Again following Diamond (1980), we differentiate the fundamental
equation of dynamic programming with respect to uE2 and uS2 : We obtain
›x2›uE2
›x2›uS2
2
6664
3
7775¼ ½ðaHp2q2 þ c2p2Þ c2p2�
r þ b2 2ðb2f2 t2Þ
2e ðr þ t2 þ eÞ
" #21
Should Policy Makers be Concerned About Adjustment Costs? 65
Solving this system of equations for ›x2=›uS2 yields
›x2›uS2
¼p2{aHq2eþ c2ðeþ r þ b2Þ}
ðr þ b2Þðr þ t2 þ eÞ þ et2 2 efb2ðB5Þ
Substituting Equation B5 into Equation B4 and collecting terms results in
rx02ðuÞ ¼p2{aHq2e2 ½ðr þ b2Þðr þ eÞ2 efb2�c2}
ðr þ b2Þðr þ t2 þ eÞ þ et2 2 efb2ðB6Þ
Note that Equation B6 is also equal to rV2TðaHÞ: Thus, since both dynamic
marginal products equal the expected lifetime income for a worker training in that
sector, and, since workers are allocated so that the expected lifetime income from
training is the same in both sectors, the dynamic marginal products are equal in
equilibrium. As a result, equilibrium is efficient.
APPENDIX C
In this section our goal is to show how to transform our measure of aggregate
income in the initial steady state into utility. In particular, we want to show that
utility is given by hðgÞYSS with
hðgÞ ¼aa½ð12 aÞð1þ gÞ�12a
aþ ð12 aÞð1þ gÞ
Given our assumption that the utility function for each consumer is given by
UðZ1; Z2Þ ¼ Za1 Z
12a2 ; it follows that the aggregate consumption bundle is given by
Z1 ¼ aI=ð1þ gÞ and Z2 ¼ ð12 aÞI; where I is a measure of aggregate income
and both world prices have been set to 1. It follows that
Z2 ¼ð12 aÞð1þ gÞ
aZ1
Now, in the tariff-distorted equilibrium, it must be the case that the value of output
equals the value of the consumption bundle when both are evaluated at world
prices. The value of output is given by YSS: It follows that Z1 þ Z2 ¼ YSS: If wenow substitute for Z2 and solve for Z1 we obtain
Z1 ¼aYSS
aþ ð12 aÞð1þ gÞ
C. Davidson and S.J. Matusz66
This implies that
Z2 ¼ð12 aÞYSS
aþ ð12 aÞð1þ gÞ
Plugging these values into the utility function then yields the desired result.
REFERENCES
Acemoglu, D. and Pischke, J. (1999). Beyond Becker: training in imperfect labor markets.
Economic Journal, 109, F112–F142.
Bacchetta, M. and Jansen, M. (2003). Adjusting to trade liberalization: the role of policy,
institutions and WTO disciplines. WTO Special Studies (#7).
Baldwin, R., Mutti, J. and Richardson, J. D. (1980). Welfare effects on the United States of
a significant multilateral tariff reduction. Journal of International Economics, 10,
405–423.
Bilsen, V. and Konings, J. (1998). Job creation, job destruction growth of newly
established, privatized and state-owned enterprises in transition economies: survey
evidence from Bulgaria, Hungary, and Romania. Journal of Comparative Economics,
26, 429–445.
Boyce, W. and DiPrima, R. (1977). Elementary Differential Equations and Boundary Value
Problems, New York: Wiley.
Brahmbhatt, M. (1997). Global Economic Prospects and the Developing Countries,
Washington, DC: The World Bank.
Cassing, J. and Ochs, J. (1978). International trade, factor market distortions, and the
optimal dynamic subsidy: comment. American Economic Review, 68, 950–955.
Davidson, C. and Matusz, S. J. (2000). Globalization and labour-market adjustment: how
fast and at what cost? Oxford Review of Economic Policy, 16(3), 42–56.
Davidson, C. and Matusz, S. J. (2001). “Globalization, employment and income: analyzing
the adjustment process,” in Trade, Investment, Migration and Labour Market
Adjustment, D. Greenaway, R. Upward and K. Wahelin (eds.), New York: MacMillan
Press, 66–92.
Davidson, C. and Matusz, S. J. (2004). An overlapping generations model of escape clause
protection. Review of International Economics (forthcoming).
Davis, S., Haltiwanger, J. and Schuh, S. (1996). Job Creation and Job Destruction,
Cambridge, MA: MIT Press.
de Melo, J. and Tarr, D. (1990). Welfare costs of U.S. quotas in textiles, steel and autos.
Review of Economics and Statistics, 72, 489–497.
Diamond, P. (1980). An alternative to steady state comparisons. Economics Letters, 7–9.
Falvey, R. and Kim, C. D. (1992). Timing and sequencing issues in trade liberalization.
Economic Journal, 102(413), 908–924.
Should Policy Makers be Concerned About Adjustment Costs? 67
Freeman, R. (1994). “How labor fares in advanced economies,” inWorking Under Different
Rules, R. Freeman (eds.), New York: Russell Sage Foundation.
Furusawa, T. and Lai, E. (1999). Adjustment costs and gradual trade liberalization. Journal
of International Economics, 49(2), 333–362.
Gaisford, J. D. and Leger, L. A. (2000). Terms-of-trade shocks, labor-market adjustment,
and safeguard measures. Review of International Economics, 8, 100–112.
Haltiwanger, J. and Vodopivec, M. (2000). Gross worker and job flows in a transition
economy: an analysis of Estonia. University of Maryland Working Paper.
Hamermesh, D. (1993). Labor Demand, Princeton, NJ: Princeton University Press.
Hamermesh, D. and Rees, A. (1988). The Economics of Work and Pay, 4th edn, New York:
Harper & Row.
Hufbauer, G. and Elliott, K. (1994).Measuring the Costs of Protection in the United States,
Washington, DC: Institute for International Economics.
Jacobson, L., LaLonde, R. and Sullivan, D. (1993a). The Cost of Worker Dislocation,
Kalamazoo, MI: W.E. Upjohn Institute for Employment Research.
Jacobson, L., LaLonde, R. and Sullivan, D. (1993b). Earnings losses of displaced workers.
American Economic Review, 83, 685–709.
Karp, L. and Paul, T. (1994). Phasing in and phasing out protectionism with costly
adjustment of labour. Economic Journal, 104, 1379–1392.
Layard, R., Nickell, S. and Jackman, R. (1991). Unemployment: Macroeconomic
Performance and the Labor Market, Oxford: Oxford University Press.
Li, J. and Meyer, W. (1996). Age as a determinant of labor’s trade policy interests. Pacific
Economic Review, 1(2), 147–168.
Ljungqvist, L. and Sargent, T. (1998). The European unemployment dilemma. Journal of
Political Economy, 106(3), 514–550.
Magee, S. (1972). The welfare effects of restrictions on U.S. trade. Brookings Papers on
Economic Activity, 3, 645–701.
Neary, P. (1982). “Intersectoral capital mobility, wage stickiness, and the case for
adjustment assistance,” in Import Competition and Response, J. Bhagwati (ed.),
Chicago, IL: University of Chicago Press.
Takacs, W. and Winters, A. (1991). Labour adjustment costs and British footwear
protection. Oxford Economic Papers, 43, 479–501.
Trefler, D. (2001). The long and short of the Canadian-U.S. Free Trade Agreement.
University of Toronto Working Paper.
C. Davidson and S.J. Matusz68
CHAPTER 4
Non-Tariff Barriers as a Test
of Political Economy Theories
PHILIP I. LEVY
Department of Economics, Economic Growth Center, Yale University,
P.O. Box 208269, New Haven, CT 06520-8269, USA
Abstract
This chapter provides a rough test of a broad and prominent class of political
economy of trade models and finds them wanting. The class features governments
with weighted social welfare functions, including the prominent model of
Grossman and Helpman. Whether the government is the single domestic player or
there are other players involved (as with the lobbies in the Grossman–Helpman
case) the government ultimately acts as a unitary player in international dealings.
Recent work has shown that such unitary actors care exclusively about terms of
trade in international negotiations. This chapter pursues the implication that
governments’ choice of trade instruments may offer a better test of the unitary
government framework than existing empirical work. We use the structure of
United States protection to argue that governments consistently choose
instruments that sacrifice terms of trade, thus casting doubt on the unitary
approach. We offer a discussion of alternative theories of political economy that
could accommodate this stylized fact.
Keywords: Trade policy, lobbying, tariffs, political economy
JEL classifications: D72, F13
E-mail address: [email protected]
4.1. INTRODUCTION
This chapter argues that the political economy of trade policy is more complicated
than most prominent models allow. In particular, the common practice of treating
governments as unitary agents in international dealings leads to a testable
implication about the choice of trade instruments: governments should pick
relatively efficient ones. In an unsophisticated test of those theories, this chapter
presents evidence that the bulk of trade protection is of a relatively inefficient form
that is inconsistent with a central prediction of the theories. There already exist a
number of sophisticated tests of these theories, but we argue that they do not
capture the essence of the models they purport to test.
The last decade saw substantial progress in theories of the political economy of
trade. To a large extent, this progress consisted of providing previously opaque
theories with solid microeconomic foundations. While theories of weighted social
welfare functions were plausible and tractable, it was not clear what kinds of
interactions among agents would give rise to them. Gene Grossman and Elhanan
Helpman’s introduction of the common agency approach to lobbying filled this
void and offered a combination that seemed almost too good to be true: a group of
heterogeneous agents involved in a lobbying competition which ultimately leads
to a single, manageable welfare function.
This rich simplicity offered the promise of performing more intricate
comparative statics or institutional analyses while retaining a political economic
foundation. That foundation was desirable because of the apparently pervasive
deviations of trade practices from standard normative trade theory such as the
prevalence of protection among countries that did not seem to have any market
power.
Even authors who might have been skeptical of the importance of political
economy found it virtually costless to include it in their models. A prominent
example is Kyle Bagwell and Robert Staiger’s theory of the General Agreement
on Tariffs and Trade (GATT). They argue that the trading system is designed to
address concerns about terms of trade movements. While that would seem to mark
a return to the days before the political economy literature flourished, Bagwell and
Staiger show that many of the prevailing political economy models also place
great emphasis on the terms of trade. In fact, they show that one can abstract away
from most political economy specifications and present a very general welfare
function in which the government cares only about the domestic price and the
international price. The important common feature of this broad class of models is
that, for a given domestic price, government welfare will increase with an
improvement in the terms of trade.
P.I. Levy70
It is this prediction that we test in this chapter. Although Bagwell and Staiger’s
model and most of the political economy theories to which they refer all deal
exclusively with tariffs, we argue that they have strong implications for the choice
of trade policy instrument. In particular, any government with this broad Bagwell
and Staiger welfare function should choose a tariff over a Voluntary Export
Restraint (VER). In general, those policies are comparable except that the VER
involves the transfer of quota rents (erstwhile tariff revenue) to the trading partner
country. This matches up very well with the theoretical prediction, since the
transfer of quota rents is equivalent to holding constant the domestic price while
worsening the terms of trade. This is precisely what governments are assumed to
dislike.
Our unsophisticated test consists of using the structure of United States trade
policy to demonstrate that the vast bulk of protection—certainly as measured by
welfare cost—takes the form of non-tariff barriers (NTBs). In fact, so prevalent
are NTBs relative to tariffs that more elaborate empirical tests of theories of tariff
levels have resorted to NTB coverage ratios as the dependent variable. These
NTBs often are structured in such a way as to worsen the terms of trade.
None of this is to argue that governments or domestic agents are acting
irrationally. Rather, the argument is that through some particular modeling
choices, the literature turned to models that featured governments as unitary
actors, even when government policy was the result of domestic competition. We
show a parallel to an older literature on household behavior, in which certain
assumptions (specifically, altruism and transfers on the part of the household
head) will lead selfish household members to work towards the common goal of
maximizing household income. Just as empirical tests of household behavior have
found instances in which households demonstrably did not act as unitary agents,
we argue that the choice of trade policy instrument has similar implications for the
unitary nature of governments.
There are a number of theoretical models that, by design or good fortune, are
compatible with the prevalence of NTBs. We attempt a taxonomy. One class relies
on the representation of foreign interests in the formation of domestic policy.
Another considers a more complex negotiating game than the one in standard
models. A third class, which we deem the most promising, features non-unitary
polities in which there is imperfect information among agents and the choice of
instrument that can send a signal.
Section 4.2 develops the linkage between weighted government welfare
functions and the choice of trade policy instrument. Section 4.3 reviews existing
empirical tests of the Grossman and Helpman framework and then uses recent US
trade policy to argue that the prevalence of NTBs that sacrifice a potential term of
trade gain constitutes a more fundamental challenge to the framework. Section 4.4
Non-Tariff Barriers as a Test of Political Economy Theories 71
considers some theoretical alternatives to the unitary framework. Section 4.5
concludes.
4.2. DEVELOPING TESTABLE IMPLICATIONS OF THE THEORY
The theory that this chapter tests is the description of governments as unitary
actors with linear social welfare functions in a competitive environment.1 In a
classic, unweighted government welfare function, the government would
maximize
WG ¼X
I
CSþX
F
PSþ TR ð1Þ
where I is the set of individuals and CS their consumer surplus, F the set of firms
and PS their producer surplus and TR the tariff revenue. If tariffs were the only
instrument, this leads to the classic policy formulation that the optimal tariff for
countries unable to affect their terms of trade is 0, while it is positive for countries
that observe an imperfectly elastic foreign export supply schedule. Of course,
even in a small country, a tariff effects a transfer from consumers to producers in
the import-competing sector, but under formulation 1 this has no beneficial effect
on government welfare.
The normative prescription for zero tariffs in small countries seemed to offer a
test for the classic formulation of government welfare as a positive theory. If small
countries had tariffs, the theory could be rejected. Indeed, the prevalence of tariffs
in small countries prompted the creation of alternative political economy
descriptions.2
A ready alternative formulation is the Stigler–Peltzman political support
function. We reinterpret that welfare function here as
WG ¼X
I
CSþX
F
PSþX
l[V
alPSl þ TR ð2Þ
1 In fact, the test applies to a somewhat broader class of models as described in detail later. We focus on
this subset for heuristic purposes since it encompasses the popular Grossman–Helpman model.2 This simple prima facie test has been challenged recently. Bagwell and Staiger (2001b, p. 79), in
arguing for the importance of terms-of-trade considerations, warn against mistaking country size for
market power. Given transport costs, “the issue is not the size of a country’s economy relative to the
world economy but rather the size of certain industries in a country relative to other industries in that
region.” Thus, in their example, Mexico would not regard Guatemala as small.
P.I. Levy72
where we have simply added a single term to Equation 1.3 That term allows
additional weight for each producer l in a set of producers with special standing
with the government, V: The weights, al . 0; can differ for different producers.
The virtue of this approach is that it offered a straightforward explanation for trade
protection: the government cares more about the recipients of transfers than those
who are being taxed. Thus, it breaks away from the uncomfortable prediction of
free trade for small countries and it does so in an eminently tractable way. The
tractability comes from the unitary nature of the government; it is easy to operate
with a single differentiable objective function. The principal disadvantage of this
approach is that it is something of a black box. We are offered no particular
insights into why the government would attach additional weight to the welfare of
firms in V; we are simply asked to believe.
It might have appeared, then, that the modeler was faced with a choice: one
could pursue “realistic” non-unitary models in which multiple agents interact,
such as the party competition models of Magee et al. (1989), but these would be
significantly less pliable than formulations such as in Equation 2. Alternatively,
one could adopt the political support function, but at the expense of any
understanding of the fundamental forces driving the model.
The seminal work of Grossman and Helpman (1994) seemed to offer a panacea.
They proposed that the government cared about two things: contributions from
lobbies and the general welfare. Thus
WG ¼ aX
I
CSþX
F
PSþ TR
!
þX
l[L
Cl ð3Þ
where Cl is the “contribution” that industry l offers the government and L is the
exogenously specified set of industries that may lobby the government in this way.
Just as the consumer surplus, producer surplus and tariff revenue terms could
fluctuate with tariff-driven price changes, the contribution could be a schedule
contingent on prices. The assumption of a prominent role for lobbying groups
matched well with descriptive work on the formation of trade policy.4 Through the
adoption of the common agency approach, the ultimate outcome of the model was
the familiar and tractable weighted government welfare function as in Equation 2.
3 More accurately, the Stigler–Peltzman approach said the government valued the rents received by
industry (Stigler’s original formulation) and consumer surplus (a Peltzman contribution). Baldwin goes
one step further and allows different weights on consumer surplus and tariff revenue. He then refers to
the result as the “deus ex machina” government objective function (Baldwin, 1987).4 See, for example, Destler (1995).
Non-Tariff Barriers as a Test of Political Economy Theories 73
This meant that one could pursue questions about trade negotiations while
enjoying solid microeconomic foundations.5
While the empirical evidence that governments value contributions is
persuasive, it is not obvious why the government should also have the general
welfare term entering linearly into its objective function. In the absence of
elections, this might be seen as a convenient and innocuous shorthand for the
broader support necessary for the government to remain in power.
Whatever the reason for the inclusion of general welfare, we can gain insight
into its effect by noting a parallel to an earlier literature on the economics of the
household. In consideration to household behavior, one prominent question that
could be asked is whether the household behaves as a single unit with a single
objective function or whether the different interests of adults and children interact
in such a way that the outcomes cannot be characterized as the maxima of a single
objective function. As an example of one such alternative, one might see a Nash
Bargaining Solution emerge from family interactions.6
In the 1970s, Gary Becker put forward his “Rotten Kid Theorem” (described in
Chapter 8 of Becker (1981)). In his setting, there are benevolent parents and
selfish children. There is a clear conflict of interests between these family
members; the parents are willing to distribute income to the children, but not to the
extent that the children would like. The theorem states: “Each beneficiary, no
matter how selfish, maximizes the family income of his benefactor and thereby
internalizes all effects of his actions on other beneficiaries.” (p. 183). Thus, under
the particular assumptions of Becker’s framework, the family’s actions can be
modeled as though it were maximizing a unitary objective function (family
income). The key to the result is that the parents are assumed to be altruistic; the
welfare of the children enters explicitly into their welfare function. The result
holds so long as the benefactor is not pushed to a corner solution (with zero
desired contributions).
Returning to the trade model of G–H, the government plays the role of
benevolent parent with altruistic concern for the other domestic agents (consumers
and shareholders in domestic firms). In their adoption of the Bernheim and
Whinston common agency framework (Bernheim andWhinston, 1986), one of the
conditions for equilibrium is that lobbies maximize the sum of their own welfare
and that of the government (by the similar reasoning that if the solution does not
maximize joint welfare, then there is additional surplus available that the lobby
could try to exploit).
5 Among the papers that did this were Grossman and Helpman (1995) and Levy (1998).6 For reviews of this literature, see Bardhan and Udry (1999, Chapter 2, Section 4) on the development
context and Bergstrom (1997) on theories of the family.
P.I. Levy74
G–H assume that governments are unable to effect the kind of direct transfers to
lobbies that Becker imagined.7 Thus, tariffs are chosen towards the same end. The
end result is a version of Equation 3 in which lobbies present the government with
“truthful” contribution schedules—ones that accurately depict the marginal effect
of a price change on profits. The government then acts as if those industries
represented by lobbies receive weight ð1þ aÞ in its objective function, while all
other agents receive weight a: Thus, we return to the weighted social welfare
function of Equation 2, but with the apparent added richness of an agency game
between competing lobbies. The point of the analogy to Becker’s model of the
family is that it is unsurprising that the government’s altruism renders the intra-
country conflicts moot when the country engages in international dealings.
In a series of recent papers, Robert Staiger and Kyle Bagwell have shown that
the unitary objective function that emerges from the G–H analysis has strong
implications for countries’ interaction in the world trading system. In particular,
they argue that such governments are principally concerned with the terms-of-
trade implications of their trade policies.8 This harkens back to the original
optimal tariff arguments of Johnson that emerged from Equation 1. Bagwell and
Staiger note that terms of trade arguments have fallen into disfavor among trade
economists, but argue that with the advent of theories such as G–Hs, they return
as a logical implication.
In a two-good model, they posit their own objective function of a very general
form
WG ¼ Wð pðt; pwÞ; pwÞ ð4Þ
where pw is the relative price of the home country’s import good (i.e., the inverse
of the terms of trade), and p represents the domestic price of the import good,
which is dependent on the world price and the tariff t (Bagwell and Staiger, 1999,p. 220). The sole restriction they place on this welfare function is
›Wðp; pwÞ=›pw , 0 ð5Þ
This specification of the welfare function has two striking effects. First,
Bagwell and Staiger show it to encompass the more specific government objective
functions of a wide range of models, including all those specified in Equations
1–3. Second, this specification clarifies the importance of the terms of trade in
those models. The restriction in Equation 5 just says that, holding the domestic
price constant, an improvement in the terms of trade makes a country better off.
7 In fact, they provide a thoughtful discussion of why lobbies might prefer the absence of direct
subsidies. We return to this point later.8 The principal work on which this discussion draws is Bagwell and Staiger (1999), but see also
Bagwell and Staiger (2001a,b).
Non-Tariff Barriers as a Test of Political Economy Theories 75
This formally captures the essence of the earlier discussion concerning the Rotten
Kid Theorem. In models of this type, though domestic agents have conflicting
interests, they interact in such a way as to be representable by a unitary objective
function. The welfare of the polity improves with better terms of trade in the same
way that the contentious members of Becker’s model family all approved of
higher income, despite their disagreements about how it should be distributed.
The central assertion of this chapter is that this yields a testable implication
for these models (and for Bagwell and Staiger’s theories of the GATT as
well). The test is an eminently basic one: if governments are principally
concerned with terms of trade in their international interactions, we should not
see them adopting instruments that forsake the terms of trade relative to other
available instruments. Specifically, we should not see a proliferation of VERs
or bilateral quota systems in which exporters control quota rights. Since these
deliver the quota rents to foreign interests, they sacrifice the terms of trade
gain that countries were ostensibly targeting through their trade policy.
Figure 4.1 depicts the classic competitive partial-equilibrium analysis of a tariff
or a VER for a large country. The initial world and domestic prices are pw0 and
imports are AD. We can consider the imposition of a tariff t ¼ ð p2 pw1 Þ=pw1 : This
raises the domestic price to p and lowers the world price to pw1 ; an improvement in
the terms of trade. Alternatively, a VER to achieve the same domestic price p
would require the trading partner to limit its exports to quantity BC. This is costly
to the home country as it reallocates the erstwhile tariff revenue (the shaded boxes)
to the foreign country.
Figure 4.1: A comparison of different trade measures in a competitive setting.
P.I. Levy76
In our context, it is important to note that the VER can readily be given a price
interpretation. Under the VER, the terms of trade will be equal to the domestic
price p; since the transfer of the quota rents constitutes an additional payment
exactly equal to the difference between the world and domestic prices. Thus, we
arrive at our testable statement
Proposition 1. Governments with objective functions as in Equations 4 and 5
should strictly prefer a tariff to a VER that achieves the same domestic price.
The proof is immediate. By assumption, the domestic price is the same in each
case. Under the VER, terms of trade are worsened and the partial derivative in
Equation 5 requires this to lower government welfare monotonically.
Note that neither Grossman and Helpman (1994, 1995) nor Bagwell and Staiger
(1999) explicitly allow for VERs nor do they make explicit predictions about
instrument choice. G–H consider tariffs exclusively, though Grossman and
Helpman (1994) features an interesting discussion of the choice between a tariff
and a more efficient subsidy. Bagwell and Staiger do raise the issue of VERs as a
means of illustrating the potential costliness of neglecting terms of trade issues.
Discussing the findings of Berry et al. (1999) on the US VER on Japanese auto
imports in the 1980s, they write:
…the decision of the United States to “give away” such an amount might
be taken as evidence that governments in fact do not care about the terms
of trade, even when the associated implications for income are large.
This inference, however, does not follow from the U.S. VER experience.
The relevant policy alternative for the United States was not a set of
unilateral tariff increases…which surely would have incited a retaliatory
“trade war” with Japan, but rather a set of tariff changes from the United
States and Japan that were consistent with GATT rules (Bagwell and
Staiger, 1999, p. 242, note 40)
This raises several points. One interpretation is that governments are playing a
game with more intricate rules that sometimes require a choice between tariffs and
VERs. If so, it would seem worthwhile to model this explicitly. A second point is
that Bagwell and Staiger note elsewhere that the GATT allows renegotiation of
previous tariff commitments. Even if renegotiation were not explicitly allowed,
the de facto remedy for the Japanese to protest US imposition of a tariff would be
either compensation or authorization to retaliate with a tariff of their own. It would
be astonishing if the net authorized damage that the Japanese could have inflicted
on the United States were of the same order of magnitude as the $8.3 billion 1983
Non-Tariff Barriers as a Test of Political Economy Theories 77
dollars estimated by Berry et al. (1999), particularly in light of the trade
imbalances that prevailed at the time.9 Third, if the only reason for adopting the
less-efficient VER is the pre-existent GATT tariff binding and this binding is
tighter in practice than it would appear, that would lead one to ask why the binding
was agreed to in the first place. If the auto VER were a unique instance, one might
believe that this was a rare case in which the government faced an unanticipated
demand for protection. In fact, as we show in Section 4.3, the use of these NTBs is
pervasive and has, at least in the case of textiles and apparel, lasted since the
1960s. Finally, if one moves beyond the case of automobiles, which were covered
by GATT agreements, one saw similar instruments employed in sectors such as
textiles and apparel and sugar which were largely outside of any GATT agreement
prior to the Uruguay Round. This casts doubt on the argument that NTBs only
serve governments as a second-best means of circumventing GATT tariff
bindings.
Before we turn to theories of political economy that may be better able to
accommodate the prevalence of NTBs, we first consider the empirical evidence.
Section 4.3 begins with a description of tests of the G–H theory that do not rely
upon instrument choice. It then considers further whether instrument choice is a
legitimate criteria for judging these model. Finally, we present stylized facts about
the relative importance of NTBs that do not improve the terms of trade.
4.3. EMPIRICAL EVIDENCE
4.3.1. Existing tests
The nature of the “test” provided in this chapter is unorthodox. It is far more
common to test a model through econometrics. This, of course, has the virtue that
a well-specified test will let us distinguish between the sort of random deviations
from a model’s predictions that are to be expected from error terms and the sort of
systematic deviations that lead us to question a theory. We begin to address this
point by considering two prominent conventional tests of the G–H model, nearly
contemporaneous papers by Goldberg and Maggi (1999) and Gawande and
Bandyopadhyay (2000). Each of these aims to confront the G–H formulation with
US data. Each is drawn to the G–H model by its more precise and rigorous
formulation, relative to its political economy predecessors. Each emerges with
9 This was the estimate of total cost of the VER over the entire period of application relative to the
estimated costs of a tariff. The imperfectly competitive framework used by Berry, Levinsohn and Pakes
to reach this estimate differed substantially from the simple competitive framework of Figure 4.1. The
point estimate of costs also had a very large standard error (also $8.3 billion, p. 401).
P.I. Levy78
a generally sanguine view of the model, in each case because the addition of
explanatory terms to the G–H formulation offered little improvement. Each paper
also qualified its empirical endorsement of the model by noting a puzzling result
or two.
Oddly enough, each also uses coverage ratios of NTBs as the dependent
variable to test this model of the level of tariff protection. Each is deliberate in this
choice and the justifications they offer are informative. Goldberg and Maggi first
argue that they avoid tariff measures because tariffs are set cooperatively (p. 1137).
They note that the same might be said of VERs; so they use one version of the
dependent variable in which only price measures, such as anti-dumping or
countervailing duties, are counted.10 They proceed to note that it might be possible
to construct a more general index of trade restrictiveness, but that this would
require data they do not have.
At any rate, we note that tariffs in the United States are very low (the
average tariff is about 5% and vary little across sectors), whereas
NTBs are higher (the average coverage ratio in our data is 13%) and
vary considerably across sectors. In addition, we suspect that
coverage ratios understate the actual extent of protection…thus, the
discrepancy between the magnitude of tariff and non-tariff protection
may be even larger (p. 1141, note 10)
Gawande and Bandyopadhyay offer a similar rationale:
The use of coverage ratios in place of what in the theory is an ad valorem
tariff requires the belief that coverage ratios are positively correlated with
their tariff equivalents across industries. The presumption becomes more
credible when, as we do, price elasticities are included to control for this
effect on the right-hand side. The computation of tariff-equivalents is an
enormously expensive task, and, given the state of the art in
computational general equilibrium, such computations are based on
assumptions about market and production structures that are merely
convenient rather than approximations to reality (p. 145, note 9)
10 It is not clear that this remedies the problem of cooperation. A non-trivial number of US anti-
dumping cases are suspended or withdrawn, for example, when the relevant parties reach agreement.
Further, the potential threat of an anti-dumping case could support a collusive outcome that has the
effect of raising a domestic price. If the threat is credible and not exercised, it is not clear that this would
show up in a coverage ratio. See, for example, Staiger and Wolak (1992).
Non-Tariff Barriers as a Test of Political Economy Theories 79
There is a curious inconsistency to these arguments. If it were true that the
fraction of subsectors covered by a NTB was sufficiently correlated with
equivalent tariff levels to allow for careful empirical testing, then the rigorous
derivations that attracted these authors to the G–H model would be unnecessary.
In fact, we have no reason to believe that the levels of tariff protection suggested
by the G–H model, dependent as they are upon terms of trade effects, should bear
any relation to the levels of drastically different instruments such as VERs or an
anti-dumping regime. It is entirely plausible that the level of protection (or even
the breadth of protection, which coverage ratios measure) afforded by these other
instruments is broadly related to factors such as import penetration or the
existence of organized political lobbies, but that simply marks a return to the
looser form of political economy realizing that preceded G–H.
Despite these obstacles, both papers support the G–H framework, in the sense
that variables which are theoretically excluded by that framework are empirically
excluded as well. Each paper has its own troubling estimate, however. Goldberg
and Maggi estimate a large weight on general welfare relative to political
contributions (0.98 vs. 0.02). They explain: “This results seems consistent with
the fact that trade barriers in the United States are quite low; even in 1983 the
average coverage ratio was only 0.13, substantially smaller than the potential
maximum of 1.” Elsewhere, however, they note that for at least one point estimate
(Goldberg, 1995), a coverage ratio of 7 percent in the auto industry corresponded
to a 60 percent tariff equivalent (p. 1141).
Gawande and Bandyopadhyay pay greater attention to estimation of the
political contribution process and ultimately express discomfort with their
estimate of government welfare weights (the term a in the G–H model). They
report that their estimate
is in conflict with the empirical evidence from computational general
equilibrium studies that have attempted to assess the welfare loss from
protection. They indicate that efficiency losses are many-fold greater
than what lobbies spend to obtain protection…Our estimates of a suggest
that (Political Action Committee) contributions are greater than dead-
weight costs, on average (p. 147)
The costly programs they cite, from which the CGE estimates are derived,
include programs in which exporting countries allocate the bilateral quota rights.
It is perhaps unsurprising that the high-efficiency costs of these programs will not
match well with the predictions of the G–H tariff model. Further, to the extent that
those programs are very costly but cover a narrow range of goods, they will
mislead estimation based on coverage ratios.
P.I. Levy80
We conclude this consideration of empirical tests with a brief mention of a third
empirical test (Maggi and Rodriguez-Clare, 2000). They deal with the same class
of models as this chapter and apply the moniker “Standard Short-run Political
Economy” (SSPE) models. They cite the G–H model as a central member of this
class. While principally focused on the estimated effects of import penetration
ratios, they argue that standard SSPE models are limited by the assumption that
governments only have access to tariffs. By contrast, “Our approach here is to
extend the standard analysis, which typically focuses on political influence by
domestic producers, to consider also political influence by foreign exporters and
domestic importers.” (p. 289).11 They consider three parameters: the political
influence of importers (not producers); the weight on foreign exporters; and the
cost of public funds. If the second parameter is highest, a VER is chosen.12
Setting aside the plausibility of foreign representation in protection decisions, it
suffices to note that this “extension” in fact marks a significant departure from the
G–H model. Specifically, while the G–H model fits comfortably within the
Bagwell–Staiger characterization of welfare, the Maggi–Rodriguez-Clare model
would not (with sufficient weight on foreign export interests, the inequality in
Equation 5 would be violated). Since this chapter argues that the basic terms of
trade prediction of that class of models fails to hold, being outside of the class is a
good thing. There are, however, a number of different theoretical approaches that
can avoid the terms of trade critique and we defer a fuller discussion of these
approaches in Section 4.4.
4.3.2. Instrument choice as a test
At a facile level, one could say that most theories of protection could be discarded
because they fail to provide a full explanation for both instrument and level. This
would be unsatisfactory since the nature of economic analysis is to simplify.
However, it is essential to check whether the omitted factors are orthogonal to the
question at hand or intimately connected. Further, there is a distinction between
theories that do not bother to explain instrument choice and those that are
11 They note that this was also the approach of Hillman and Ursprung (1988).12 There is recent work arguing that foreign lobbying is a significant factor in the determination of trade
policy in the United States (Gawande, Krishna and Robbins, 2002). It uses data on payments to “foreign
agents” in the United States as a proxy for otherwise prohibited direct political contributions from
foreigners. There is reason to doubt, however, that these agents (lobbyists) are being used as conduits
for funds as opposed to providing general representation. It is not clear what role there would be for
non-pecuniary representation in a G–H model. Such an approach would seem to fit better with a model
in which information played an important role.
Non-Tariff Barriers as a Test of Political Economy Theories 81
intrinsically incapable of being extended to explain it. To consider whether
confronting the selected theories with NTBs constitutes a fair test, we briefly turn
to the literature on instrument choice.
In a seminal paper on the topic Becker (1983) puts forward an argument
very much in keeping with the class of models this chapter critiques. He
considers government policies more generally than just those on trade and
argues that instruments will be chosen in an efficient fashion. The rationale is
that there will be a surplus generated by the move from a less efficient to a
more efficient instrument and that this will either present a direct additional
gain to lobbying groups or diminish opposition making their original goal
more attainable.
There is an interesting qualification that Becker offers to his broad assertions
about efficiency. He writes
I have assumed that influence functions depend only on the character-
istics of and the pressures exerted by political groups, and not on taxes
and subsidies, the number of persons in each group, the distribution of
income, or other variables. The ignorance of voters not only helps
determine the influence of different characteristics and pressure, but may
also make influence depend on other variables as well. For example,
influence may depend on subsidies if voters mistakenly believe that
certain subsidies (minimum wages or oil entitlements?) contribute to
desired goals rather than to the incomes of particular groups. If influence
functions were affected by taxes, subsidies, and other policies, the
analysis in this paper might have to be significantly modified, including
the conclusion that efficient taxes tend to dominate inefficient taxes…or
that policies raising efficiency tend to have greater political support than
policies lowering efficiency… (p. 394)
That will, in fact, be the class of explanation favored in this chapter. There is
imperfect information among voters and they draw different signals from VERs
than they do from tariffs.
Becker offers a subtle interpretation of efficiency which insulates him
somewhat against the sort of prima facie evidence that this chapter offers.
Instruments are more efficient if the net outcome is more efficient. Thus, a tariff
could be more efficient than a direct production subsidy if few users were able
to take advantage of a tariff scheme while a subsidy scheme would be widely
available. The deadweight losses would be greater for each instance of a tariff,
but the decrease in the number of instances relative to subsidies would offset
the cost.
P.I. Levy82
In the trade policy literature, attention has mostly focused on the question of
why tariffs are used in lieu of more efficient subsidies (Rodrik, 1995, Section 4).
Surveying a number of models that compare equilibria under different sets of
policy instruments, Rodrik writes
In each of the…models, the comparison involves equilibria of
different “policy regimes”, where each regime is characterized by the
use of a specific policy (tariffs or production subsidies, say). What is
often left vague is the political mechanism that governs the choice of
one regime over another. One can think of this choice as being made
in the first stage of a two-stage political economy model. This
appears to be the implicit view in the previous papers, but the
decision-making process for this first stage is not well specified in
any one of them (p. 1473)
Grossman and Helpman’s (1994) work is among those that Rodrik surveys. It
includes a section arguing that lobbies might well prefer tariffs to more direct
subsidies since competition could be more fierce in the latter case. Implicit in the
discussion is that the challenge to the model is likely to come from more efficient,
not less efficient, instruments. Also, as Rodrik notes, the question is treated as
separable.
It may be worth asking whether it is fair to subject theories that set out to
explain tariffs to a test on their ability to explain VERs. We contend that it is.
Grossman and Helpman are describing “Protection for sale”, not “Tariffs for
sale”. Bagwell and Staiger are offering a general description of the
governance of the world trading system and asserting that terms of trade
effects are central.
One should be able to neglect an explanation of NTBs under three possible
assumptions. First, one might argue that conclusions that are drawn in the realm
of tariffs can be readily extended to NTBs. A central purpose of this chapter
is to argue that this is generally not so. Second, it might be that the complications
that exist in determining the levels of NTBs are readily separable from tariff
determination, so the two can be examined separately. That has been standard
practice, but it is hard to justify why the same polity would give primacy to terms
of trade in tariff determination and willfully neglect it in determination of VERs.
If VERs are adopted because tariffs are bound, then the determination of the
levels of each instrument is not independent. Finally, one could assume that
NTBs are a relatively minor exception to the standard practice of tariffs. We
show in Section 4.3.4 that the reverse is a better description of the structure of US
trade policy.
Non-Tariff Barriers as a Test of Political Economy Theories 83
4.3.3. Evidence on non-tariff barriers
To the best of our knowledge, there does not exist any estimate that divides the
totality of US protection into different trade policy instruments. In fact, any
attempt to do so would face some immediate obstacles. First, there is the
omnipresent question of how one would weight the different instruments. If VERs
covered 5% of tariff lines, accounted for 20% of consumer surplus loss and 50% of
welfare costs, which figure could one use to assess their importance? We address
this by offering multiple measures from secondary sources below. A second
measurement question concerns more complicated instruments such as anti-
dumping policy or rules of origin. To the extent that the threat of a dumping action
deters competitive pricing on the part of a foreign exporter, this will have the
welfare effects of a VER but will be difficult to measure. Rules of origin in
preferential trade agreements have been shown to afford protection to producers of
intermediate goods within the PTA (see Krueger and Krishna, 1995). We are
unable to overcome this difficulty, which is equally problematic for the
measurement of levels and for coverage ratios; one can only make rough guesses
at the extent of bias. Finally, if we are to offer a “test” of the existing theories, it is
worth considering what standard we might use to determine the power of the test.
Given the bold prediction in Proposition 1 that countries with the objective
functions of Equation 4 or 5 should never choose a VER or related instrument over
a tariff, one might conclude that any sighting of a VER would constitute a
rejection.13 However, without explicitly introducing an error term, it is desirable
to allow for some aberrant behavior. This would pose a serious concern if NTBs
were a minor element of US protection. Given their prominence, we simply assert
that while the cutoff line for rejecting the proposition is difficult to place precisely,
it lies somewhere well short of the evidence.
To show this, we consider several sources on the structure of US protection.
First, and most recent, there is the official review of the United States under the
World Trade Organization’s Trade Policy Review Mechanism (WTO, 1999). It
finds that the US economy is generally open, particularly with regard to tariffs
Most imports either enter the United States duty free or are subject to
very low tariffs, all except two of which are bound. Zero tariffs apply to
nearly one third of national tariff lines and the simple average MFN tariff
rate has declined from 6.4% in 1996 to 5.7% in 1999; the average can be
13 There is an interesting discussion of this general point at the beginning of Leamer and Levinsohn
(1995), who grapple with reconciling the abstractions of trade theory with the realities of data and
whether one should take the theory too seriously or treat it too casually.
P.I. Levy84
expected to fall to 4.6% once the Uruguay Round and (Information
Technology Agreement) tariff cuts are fully implemented (p. xxi)
There are, however, tariff peaks that are substantially higher in sectors such as
agriculture and food products, textiles, clothing and footwear.14
The report notes, without quantifying, the existence of bilateral quotas,
particularly for textiles and apparel. It also documents 742 anti-dumping
investigations between 1980 and 1998, of which 44% resulted in final
affirmative findings (p. 67). From this, we can take only that a range of
instruments are applied, without any particular sense of their relative
magnitude.
Robert Feenstra provides a survey of estimates of costs stemming from US
import protection in the mid-1980s (Feenstra, 1992). Blending partial
equilibrium and general equilibrium results, he estimates the total cost of
US protection to lie between $15 billion and $30 billion, compared to 1985
US GNP of $4 trillion (p. 166). He goes on to argue for reasons why this
range might underestimate the costs of protection. For our purposes, though, it
is interesting to note the decomposition of those costs. The surveyed studies
suggested that between $7.9 billion and $12.3 billion of the costs were due to
US deadweight loss, while $7.3 billion to $17.3 billion were due to the loss of
quota rents. This offers one measure of the relative importance of VERs and
related instruments—they account for roughly half the cost of protection as
compared to the costs that would be incurred with tariff equivalents. These
quota rent transfers are the ones that are not supposed to exist at all under the
objective functions we are considering. We get an even more striking measure
if we compare the total cost of VERs to the total costs of tariffs. This latter
figure is estimated to be between $1.2 billion and $3.4 billion, thus leaving
roughly 90% of the cost of protection as attributable to a VER or related
instrument.
In a separate study, Gary Hufbauer and Kimberly Elliott provide general
equilibrium estimates of the costs of US protection in 1990 (Hufbauer and Elliott,
1994). They focus on 21 cases that, they say, account for roughly half the national
net welfare cost of protection. They report that
14 Interestingly, in light of arguments about how well informed domestic agents might be about
policies’ impacts, the WTO Secretariat notes that only one in seven duties are specific, but that
specific duties account for 86 of the top 100 MFN tariffs (p. xxi). Their interpretation is that
“such duties are intrinsically more opaque than ad valorem duties and can be used to conceal high
ad valorem equivalents (AVEs).” (p. 48). They do note that the US regularly publishes AVEs for
its specific duties.
Non-Tariff Barriers as a Test of Political Economy Theories 85
The net national welfare gain from liberalization in these sectors amounts
to an estimated $10 billion, with more than two-thirds being quota rents
recaptured from foreign exporters and producers, mostly in the textile
and apparel sectors (p. 7)
In another version of their model, they allow for changes in the terms of trade
and set the world supply elasticities for exports to the United States at 3.0, a value
they deem conservative. They then calculate net national welfare changes
incorporating increases in import prices. There are a range of cases for which
liberalization would entail a terms-of-trade loss and would therefore provide
support for the class of theories discussed in Section 4.2.15 For these cases the
terms-of-trade losses totaled $538 million, while there was an estimated net
welfare loss from liberalization of $426 million. For cases in which potential
terms of trade losses were negated by the transfer of quota rents to foreigners,16
the quota rents reclaimed under liberalization would total $2993 million, while the
total welfare gain from liberalization would be $9374 million.
Thus, the theoretical implication of the class of models considered in Section
4.2 was that in international dealings, terms of trade should be of paramount
importance. Instead, we see that for the most part, US trade policy can be
characterized as giving away quota rents. As noted above, given the transparency
of tariffs and the opacity of some measures of protection that were not included in
these studies, they probably understate the relative importance of instruments that
forsake the terms of trade. Presumably, the rents are being foregone to achieve
some other goal. In Section 4.4 we offer a brief description of models that could fit
the stylized fact of pervasive NTBs.
4.4. ALTERNATIVE EXPLANATIONS
There are a number of papers in the literature that either explicitly allow for VERs
or offer promising approaches to political economy that might justify them. We do
not attempt to argue for one over the other, only to group them into broad
categories.
15 The sectors are ball bearings, benzenoid chemicals, canned tuna, ceramic articles, ceramic tiles,
costume jewelry, frozen concentrated orange juice, glassware, luggage, polyethylene resins, rubber
footwear, softwood lumber, women’s footwear (except athletic), women’s handbags and dairy
products. (p. 28).16 These were sugar, apparel, textiles and machine tools. The estimates for apparel dwarf the other
cases ($2413 million in foregone quota rents and $7712 million in total gain from liberalization).
P.I. Levy86
The first category is the one in which foreign interests explicitly enter into
the government objective function. Maggi and Rodriguez-Clare (2000),
discussed earlier, allows the government to place some weight on foreign
interests and finds that if the weight is sufficiently high, a VER may be
adopted. Hillman and Ursprung (1988) present a model in which candidates
compete for elective office and the probability of election increases with
campaign contributions. Both domestic and foreign producer interests are
allowed to make these contributions. They find that tariffs will be a divisive
policy while VERs are not. Thus, contributions will be higher under VERs
and candidates will prefer them to tariffs. The advantage of this approach is
that one can use more conventional modeling formulations. The disadvantage
is that it is unclear why foreign interests should enter directly into government
objective functions. One might imagine that direct contributions from foreign
interests would either be illegal or impart a stigma. It seems more likely that
the weight attached to foreign interests is a reflection of constraints imposed
through threats of retaliation, for example.
The second category consists of models in which the more complicated
structure of international negotiation helps shape instrument choice. Feenstra and
Lewis (1991) use a weighted social welfare function of the sort in Equation 2.
However, they also posit international negotiations in which the foreign country is
uncertain about the extent of domestic political pressure on the home country.
They show how VERs may be preferable to tariffs in their ability to get the home
country to accurately represent the pressures it faces for protection. In a model
with imperfect competition Rosendorff (1996) incorporates VERs and anti-
dumping policy. Firms are engaged in Cournot competition, which undoes some
of the puzzle about the use of VERs. As modeled, domestic firm profits rise with
the VER relative to the optimal tariff; so governments may prefer VERs.
Rosendorff contrasts tariffs and VERs since these are the common outcomes of
anti-dumping cases. However, a contrast between quotas and VERs might be more
apt in a model of quantity competition. As in Feenstra and Lewis, the foreign
government is uncertain about home country characteristics. Home is explicitly
playing a game against foreign and Rosendorff solves for the Perfect Bayesian
Equilibrium.
The third and final category consists of models in which domestic imperfect
information plays an important role. While the models in the second category
featured imperfect information on the part of the partner country Coate and Morris
(1995) feature domestic voters who are uncertain whether the politicians in office
are good or bad. The politician’s choice of instrument sends a signal about his
type. Though this model does not explicitly address the determination of trade
policy, it demonstrates a mechanism whereby political economy plays a
Non-Tariff Barriers as a Test of Political Economy Theories 87
substantially richer and more important role than it does under the Bagwell and
Staiger interpretation.
4.5. CONCLUSION
This chapter has argued that the central paradigm of recent theories of the political
economy of trade policy is in conflict with the central feature of trade policy as
practiced—a heavy reliance on instruments that forsake terms-of-trade gains,
presumably in order to achieve domestic transfers. This has important
implications not just for our understanding of trade policy formation, but also
for our understanding of institutions such as the World Trade Organization that
tries to facilitate policy coordination.
Perhaps the most difficult aspect of modeling the political economy of trade is
finding an interior solution. It is relatively easy to posit the existence of lobbies or
other forces that have disproportionate power to sway policy. It is difficult to
explain why these groups may be able to obtain only part of what they seek. One
prominent attempt to solve this puzzle—the inclusion of additive general welfare
in the Grossman and Helpman model—resulted in the prediction of a government
that acted in unitary fashion and sought efficient policies.
The empirical evidence casually reviewed here suggests that opaque inefficient
policies may be easier to implement than transparent and efficient ones. This
argument is not new; Brock, Magee and Young described this as the principle of
“optimal obfuscation.” However, it still lacks firm theoretical underpinnings. The
provision of such a theoretical foundation should allow the sort of rigorous
analysis of trade policy that the Grossman and Helpman approach promised.
ACKNOWLEDGEMENTS
I am grateful to Stephen Morris, Chris Udry and T.N. Srinivasan for very helpful
discussions. All remaining errors are my own.
REFERENCES
Bagwell, K. and Staiger, R. W. (1999). An economic theory of GATT. American Economic
Review, 89(1), 215–248.
Bagwell, K. and Staiger, R. W. (2001a). Domestic policies, national sovereignty, and
international economic institutions. Quarterly Journal of Economics, 116(2), 519–562.
P.I. Levy88
Bagwell, K. and Staiger, R. W. (2001b). The WTO as a mechanism for securing market
access property rights: implications for global labor and environmental issues. Journal of
Economic Perspectives, 15(3), 69–88.
Baldwin, R. (1987). Politically realistic objective functions and trade-policy PROFs and
tariffs. Economics Letters, 24(3), 287–290.
Bardhan, P. K. and Udry, C. (1999). Development Microeconomics, New York: Oxford
University Press.
Becker, G. S. (1981). A Treatise on the Family, Cambridge, MA: Harvard University Press.
Becker, G. S. (1983). A theory of competition among pressure groups for political
influence. Quarterly Journal of Economics, 98(3), 371–400.
Bergstrom, T. C. (1997). “A survey of theories of the family,” in Handbook of Population
and Family Economics, M. R. Rosenzweig, and O. Stark (eds.), New York: Elsevier,
21–79.
Bernheim, B. D. and Whinston, M. D. (1986). Common agency. Econometrica, 54,
923–942.
Berry, S., Levinsohn, J. and Pakes, A. (1999). Voluntary export restraints on automobiles:
evaluating a trade policy. American Economic Review, 89(3), 400–430.
Coate, S. and Morris, S. (1995). On the form of transfers to special interests. Journal of
Political Economy, 103(6), 1210–1235.
Destler, I. M. (1995). American Trade Politics, Washington, DC: Institute for International
Economics, Twentieth Century Fund.
Feenstra, R. C. (1992). How costly is protectionism. Journal of Economic Perspectives,
6(3), 159–178.
Feenstra, R. C. and Lewis, T. R. (1991). Negotiated trade restrictions with private political
pressure. Quarterly Journal of Economics, 106(4), 1287–1307.
Gawande, K. and Bandyopadhyay, U. (2000). Is protection for sale? Evidence on the
Grossman–Helpman theory of endogenous protection. Review of Economics and
Statistics, 82(1), 139–152.
Gawande, K., Krishna, P. and Robbins, M. (2002). “Foreign lobbies and US trade policy
(mimeo)”.
Goldberg, P. K. (1995). Product differentiation and oligopoly in international markets–the
case of the United-States automobile-industry. Econometrica, 63(4), 891–951.
Goldberg, P. K. and Maggi, G. (1999). Protection for sale: an empirical investigation.
American Economic Review, 89(5), 1135–1155.
Grossman, G. M. and Helpman, E. (1994). Protection for sale. American Economic Review,
84(4), 833–850.
Grossman, G. M. and Helpman, E. (1995). Trade wars and trade talks. Journal of Political
Economy, 103(4), 675–708.
Hillman, A. L. and Ursprung, H. W. (1988). Domestic politics, foreign interests, and
international-trade policy. American Economic Review, 78(4), 729–745.
Hufbauer, G. C. and Elliott, K. A. (1994). Measuring the Costs of Protection in the United
States, Washington, DC: Institute for International Economics.
Non-Tariff Barriers as a Test of Political Economy Theories 89
Krueger, A. and Krishna, K. (1995). “Implementing free trade areas: rules of origin and
hidden protection,” in New Directions in Trade Theory, J. Levinsohn, A. V. Deardorff
and R. M. Stern (eds.), Ann Arbor, MI: University of Michigan Press.
Leamer, E. E. and Levinsohn, J. A. (1995). International trade theory: the evidence, vol. 3,
Handbook of International Economics, G. M. G. a. K. Rogoff (ed.), New York: North-
Holland, 1333–1394.
Levy, P. I. (1998). Free Trade Agreements and Inter-Bloc Tariffs, New Haven, CT:
Department of Economics, Yale University.
Magee, S. P., Brock, W. A. and Young, L. (1989). Black Hole Tariffs and Endogenous
Policy Theory: Political Economy in General Equilibrium, Cambridge: Cambridge
University Press.
Maggi, G. and Rodriguez-Clare, A. (2000). Import penetration and the politics of trade
protection. Journal of International Economics, 51(2), 287–304.
Rodrik, D. (1995). Political Economy of Trade Policy, vol. 3, Handbook of International
Economics, G. M. G. a. K. Rogoff (ed.), New York: North-Holland, 1457–1494.
Rosendorff, B. P. (1996). Voluntary export restraints, antidumping procedure, and domestic
politics. American Economic Review, 86(3), 544–561.
Staiger, R. W. and Wolak, F. A. (1992). The Effect of domestic antidumping law in the
presence of foreign monopoly. Journal of International Economics, 32(3-4), 265–287.
WTO (1999). Trade Policy Review: The United States, Geneva: The Organization.
P.I. Levy90
CHAPTER 5
The Peculiar Political Economy of
NAFTA: Complexity, Uncertainty
and Footloose Policy Preferencesw
H. KEITH HALLa and DOUGLAS R. NELSONb,c,*
aUS Department of Commerce, 14th and Pennsylvania Avenue, Washington, DC 20230, USAbMurphy Institute of Political Economy, 108 Tilton Hall, Tulane University,
New Orleans, LA 70118, USAcLeverhulme Centre for Research on Globalisation and Economic Policy, School of
Economics, University of Nottingham, University Park, Nottingham NG7 2RD, UK
Abstract
On a large number of dimensions, the domestic political economy of the North
American Free Trade Agreement (NAFTA) in the US was peculiar. In some
ways the most surprising aspect of the politics of NAFTA relates to the apparent
footlooseness of aggregate opinion. That is, since standard theories of political
economy assume that policy preferences are determined by material conditions,
those theories only predict changes when material (economic or political)
conditions change. In this chapter we provide evidence that aggregate
public opinion in NAFTA shifted dramatically in the absence of any change
in the underlying political and economic fundamentals. We will then sketch
w This chapter is solely meant to represent the opinions of the authors, and is not meant to represent in
any way the views of the US Government or the Department of Commerce.*Corresponding author.
E-mail address: [email protected]
the elements of a theory of footloose policy preferences that helps understand
this sort of phenomenon and conclude with a discussion of the implications of
the analysis for policy analysis and advice.
Keywords: Political economy, trade, NAFTA, learning
JEL classifications: D72, D83, F13
On a large number of dimensions, the domestic political economy of the North
American Free Trade Agreement (NAFTA) in the US was peculiar. For example,
at least from the perspective of standard endogenous policy models of trade
policy, the coalitions that emerged would seem to be difficult to rationalize in an
ex ante compelling way.1 Similarly, the level of national political heat generated
by a policy estimated to have only very modest effects is surprising. However, in
some ways the most surprising aspect of the politics of NAFTA relates to the
apparent footlooseness of aggregate opinion. That is, since standard theories of
political economy assume that policy preferences are determined by material
conditions, those theories only predict changes when material (economic or
political) conditions change. In this chapter we will argue that aggregate public
opinion in NAFTA shifted dramatically in the absence of any change in the
underlying political and economic fundamentals. We will then sketch the
elements of a theory of footloose policy preferences that helps understand this sort
of phenomenon and conclude with a discussion of the implications of the analysis
for policy analysis and advice.2
1 We specifically have in mind the strong opposition of a number of unions in sectors where research
suggested that NAFTA would permit rationalization of production involving increased production in
the US. The most obvious case is the automotive sector where an end to domenstic content rules will
permit rationalization of production that was estimated to increase US employment in that sector.
Similarly with respect to environmental organizations where research was far from clear on the
environmental effects of NAFTA (though here it is important to recall that environmental organizations
were actually split on NAFTA). With respect to bizarre coalitions, we need only recall that the
Democrat president and the Republican congressional leadership led one coalition, opposing a group
led by Pat Buchanan, Ross Perot, Richard Gephardt, Jesse Jackson, and Ralph Nader. Note that this is
not to say that such a rationalization cannot be produced. We have no doubt that several can. We are,
after all, professionals. The question is whether such a rationalization can be produced that would have
been compelling ex ante.2 The theory of footloose policy preferences is developed in detail in Nelson (1998).
H.K. Hall and D.R. Nelson92
5.1. THE FACT: FOOTLOOSE AGGREGATE PREFERENCES
ON NAFTA
In this section we first provide evidence of a sizable shift in public opinion on
NAFTA and then provide evidence that political and economic fundamentals
were, at best, unchanged and, at least arguably, moved in such a way that would
have generally been expected to increase support for NAFTA.3 Figure 5.1 shows,
from 1991 to 1995, the shares of responses to questions asking whether the
respondent believes NAFTA “would be mostly good or mostly bad for the US”
and, from 1996 to 2000, whether NAFTA has “had more of a positive impact or
more of a negative impact on the US”.4 It is easy to see that from very high levels
of positive evaluation and very low levels of negative evaluation in 1990 and
1991, the positives drop dramatically while the negatives rise more slowly, but
still steadily, so that by the mid-1990s positive and negative evaluations are
essentially balanced.5 Figure 5.2 summarizes these two by taking the ratio of
“good” to “bad” responses. We now want to argue that this shift in opinion occurs
despite the fact that, although NAFTA had been officially “implemented” in
January 1994, virtually nothing of economic substance had occurred, or was
scheduled to occur for several years. Furthermore, we want to argue that nothing
had changed in Mexican or US fundamentals that would lead to a policy
reassessment of the sort revealed by these two polls.
First, NAFTA itself. In early 1990, as part of, and as a support to, extensive
domestic economic reforms, Mexico’s president (Carlos Salinas de Gortari)
approached the Bush White House with a request to negotiate a bilateral free trade
agreement with the United States.6 Canadian concern with the implications of
such an agreement for its recently negotiated trade agreement with the US led to
3 Philip Levy (1998) discusses such a shift in terms of two polls asking: will NAFTA be “mostly good
for the country” (1991); and has NAFTA had “more of a postive…or a negative impact” (1996). It turns
out that these two polls are the most extreme polls on NAFTA in the Roper Center database.4 These data are drawn from the Roper Center for Public Opinion Research database (available from
Lexis–Nexis), and include all questions in that database of the form described in the text. The number
of polls in each year are 1990 (1); 1991 (2); 1992 (3); 1993 (3); 1994 (2); 1995 (0); 1996 (2); 1997 (8);
1998 (1); 1999 (3); and 2000 (1).5 Following a sizable jump, from 1 to 19%, between 1990 and 1992, undecided responses stay in a
fairly narrow range around 20% (except for 1999 (32%)). Null responses—i.e., either “about the same”
or “neither one nor the other”—also make up a small number of responses (usually less than 5%).6 It is useful to recall that Mexico acceded to GATTmembership in 1986. This was part of a substantial
trade liberalization begun by President de la Madrid in 1983. See Ten Kate (1992), Weiss (1992),
Pastor and Wise (1994), and Tornell and Esquivel (1997) for useful discussions of the trade policy
aspects of Mexican liberalization in the period leading up to NAFTA. For more on Mexican reforms
generally, see Lustig (1998).
The Peculiar Political Economy of NAFTA 93
the continental approach that eventually produced NAFTA. Virtually from the
beginning, labor and environmental interests began organizing against NAFTA.
The central issue in the early period was fast-track negotiating authority and,
while there was substantial public opposition by labor and environmental groups,
but consistent with the poll evidence cited above, fast track passed with substantial
majorities on 23 (House) and 24 (Senate) May 1991.7 NAFTA then became a
major public issue in the Presidential campaign. However, only minor candidate
Ross Perot opposed NAFTA (both George Bush and Bill Clinton came out
publicly in support).8 Perhaps surprisingly, NAFTA passed with nearly the same
margins as had fast track, though in this case the outcome was considered highly
uncertain until virtually the moment of the vote.9 Under this legislation, NAFTA
was scheduled to be implemented in January 1994.
One of the tricky things about implementation of complex legislation like
NAFTA is the timetable. The NAFTA was completed in August 1992
and implementing legislation in the US became effective on January 1, 1994.
Figure 5.1: Positive (B) and negative (X) evaluations of NAFTA.
7 231–192 in the House and 59–36 in the Senate. These majorities are actually majorities against
disapproval resolutions. These counts are from Destler (1995), which provides a fine treatment of the
politics surrounding NAFTA, and all other aspects of US trade policy. Boadu and Thompson (1993)
and Kahane (1996a,b) provide conventional econometric studies of the fast track vote.8 It should be noted, however, that candidate Clinton did express concern about both labor and
environmental issues, stressing the importance of side agreements on both.9 House, 234–200 (17/11/93); and Senate, 61–38 (20/11/93). The NAFTA votes, especially in the the
Senate, have been extensively studied, see Conybeare and Zinkula (1994), Steagall and Jennings
(1996a,b), Kahane (1996a,b), Thorbecke (1997), Holian et al. (1997), Kamdar and Gonzalez (1998),
Bailey and Brady (1998), and Baldwin and Magee (2000).
H.K. Hall and D.R. Nelson94
It incorporated and expanded most of the provisions of the CFTA that was halfway
into its 10-year phase-in period. While the agreement eliminated tariffs on many
goods immediately (over half of US imports and nearly a third of exports with
Mexico), it began the phasing-out of remaining tariffs over a 15-year period. Most
of the anticipated effects of the agreement on the US economy, however, were
through provisions covering a broad range of nontariff barriers, foreign direct
investment, intellectual property rights, trade in services, and a number of other
trade facilitating agreements (on customs administration, product standards,
antitrust, and telecommunications). Many of these provisions were also to be
phased-in over time. There were also two well-publicized side agreements on
environmental and labor cooperation.
Prior to NAFTA implementation, the average US tariff rate on imports from
Mexico was only around 3% and half by value already entered the US duty-free.
By 1996, about three-fourths of US imports from Mexico were duty free with an
average tariff of around 2.5% on the remainder (including sectors such as motor
vehicles and motor vehicle parts, apparel and textiles, and fresh vegetables). The
average tariff rate in Mexico on US goods started at around 10% prior to NAFTA
and in 1996 was down to approximately 3% on NAFTA goods. Also, the share of
US exports entering Mexico duty free increased from roughly one-third to two-
thirds by 1996.
Most provisions on nontariff barriers were still being implemented in 1996. In
particular, prohibitions, quantitative restrictions, and import licensing require-
ments by Mexico were still being phased-out with the use of tariff-rate quotas.
Additionally, implementation was not yet complete in a number of key industries.
For example, Mexico is phasing-out trade and investment restrictions on
Figure 5.2: Ratio of positive to negative evaluations.
The Peculiar Political Economy of NAFTA 95
automobiles over 10 years—including a slow reduction in local content
requirements from 36% to only 34% in the first 5 years. Additionally, the US is
phasing-out quotas on textiles and apparel made with foreign material over a 10-
year period and both countries are phasing out nontariff barriers on agriculture
(import licenses in Mexico and quota shares in the US) over 10 or 15 years using
tariff-rate quotas. Implementation by Mexico is also not complete in the
telecommunications, transportation services, and financial services industries. The
provisions related to foreign direct investment, intellectual property rights, trade
in services, customs administration, and product standards were fully in effect
by 1996.10
Complicating an assessment of the impact of NAFTA on the US was the fact
that theWTOAgreements entered into effect only 1 year later and addressed many
of the same issues. In general, the WTO Agreements were broader than the
NAFTA and went farther in sectors such as agriculture and telecommunication
services, but fell short with respect to foreign direct investment and government
procurement. Notable overlap includes areas such as sanitary measures, textile
and clothing, antidumping, safeguards, intellectual property rights, and dispute
settlement. The WTO Agreements also lowered US tariffs by nearly a third and
began a 5-year phase-in period of tariff reductions on most products, with a 10-
year phase-in period on sensitive sectors (such as textiles). Mexico did not reduce
their MFN tariffs as a result of the WTO Agreements and, instead, merely bound
their rates generally at 35%. In fact, in response to the peso crisis, Mexico raised
MFN tariffs on 502 consumer goods in 1995 from an average of 20–35%. US
exports to Mexico under the NAFTA were exempt from these tariff increases.
Thus, the major political event in the period 1991–1992 was the fast-track vote.
The passage of the legislation, and thus the final legal form of NAFTA was not
determined until the end of 1993, with “implementation” occurring in 1994.11 It
would be hard to argue that the sizable shift in opinion occurring between 1991
and 1993 was a function of the sort of shifts in political fundamentals that might be
taken to account for changed policy preferences in standard political economy
models. While the institutional environment had not changed significantly, it is
entirely possible that a change in economic conditions induced the change in
public trade policy preferences. We know from previous research on the correlates
of trade policy that trade balance and general macroeconomic conditions are
10 For a much more detailed discussion of the status of NAFTA implementation in 1996, see Chapter 2
of The Impact of the North American Free Trade Agreement on the US Economy and Industries: A
Three-Year Review, US International Trade Commission, publication 3045.11 With reference to Levy’s (1998) paper we simply note that (1) opinion shifts well before this date;
and (2) actual implementation occurs much later (i.e., many of the most significant changes in US and
Mexican law were scheduled to occur only with lags of up to 15 years).
H.K. Hall and D.R. Nelson96
associated with changes in trade policy demands. Figure 5.3A shows the balance
on goods and services (1980–1999), where 1991 is the last year of a multi-year
improvement in the trade balance and where 1992 and 1993 show positive
performance by historical standards. By contrast, the large deteriorations in 1998
and 1999 do not appear to be particularly associated with changed evaluations.
Similarly, Figure 5.3B shows the (possibly more relevant) merchandise trade
balance with Mexico, where we see 1991 and 1992 are years of improving trade
balances and, again, the large deterioration from 1994 to 1995 is not associated
with changed evaluations. Aggregate economic indicators, shown in Figure 5.4,
are equally unsupportive of an economic fundamentals story.12 The period of
Figure 5.3: (A) US balance on trade in goods and services. (B) US–Mexico merchandise trade
balance.
12 It should also be recalled that this is the period of realignment of the US dollar from a period of very
large overvaluation. It will be recalled that the dollar began accelerating in late 1980 or early 1981,
reaching its peak in February 1985, and ultimately returning to something like an equilibrium level in
late 1987 or early 1988. The essential fact would seem to be that 1991 is at the end of a period of
exchange rate instability, while 1996 is in a relatively stable period.
The Peculiar Political Economy of NAFTA 97
strongest support coincideswith a period of low (even negative in 1991) growth and
high unemployment, while the periods of weaker support are characterized by
stronger growth performance and lower unemployment. The picture that emerges
from these data is clear: standard theoretical and empirical models of trade policy
preference/behavior would have predicted greater activism in 1990–1993 than in
the latter half of the decade. Interestingly, this suggestion is supported by the data in
Figure 5.5, which shows the number of Title VII (anti-dumping and countervailing
duty) petitions initiated per year. Activity in the administered protection
mechanisms, of which the Title VII mechanism is the most prominent, is well
known to be a major indicator of protectionist activity. That figure shows a
relatively smoothly rising trend in filings consistent with a period of deteriorating
economic performance, and a drop in 1995 to a lower annual level offilings.13Thus,
as we asserted at the outset of this section, it would be difficult to attribute to
changed institutional ormaterial conditions the shift in public attitudes with respect
to NAFTA.
Figure 5.4: Aggregate economic indicators: (P) unemp; (X) drlGDP.
13 The spikes in 1992 and 1997 are associated with massive steel industry filings that are determined
institutionally (i.e., as a part of the steel industry’s strategy relative to the mechanism) and are not in
any obvious way associated with Mexico, Canada, or NAFTA.
Although, using annual data (1980–1999 and 1990–1999) there are too few data points for
convincing analysis, simple regression of Title VII petitions on unemployment and trade balance
suggests that both of these play a role in explaining filing behavior (with unemployment playing the
larger role). Carrying out the same analysis using positive responses, negative responses, or their ratio
yields no significant results at all.
H.K. Hall and D.R. Nelson98
5.2. POLICY COMPLEXITY, SOCIAL LEARNING
AND FOOTLOOSE PREFERENCES
As economists/political-economists, the conclusion of Section 5.1 leaves us in
an awkward position. All positive political economy models operate by
assuming that the policy preferences of individuals are derived in some
relatively straightforward way from the effect of policy (and policy changes) on
material well-being. But we have just displayed a case where policy preferences
changed, and changed dramatically, with no relevant change in material or
institutional conditions. In this section we argue that the complexity of NAFTA
makes determining the effects of implementation difficult. It is under precisely
such conditions that we might expect agents to condition their behavior on the
actions of others, not for strategic reasons, but because those who have already
taken actions might reasonably be expected to possess knowledge one does not
possess. This attempt to learn from the behavior of others is called social
learning.
From the point of view of the citizen, or policy-maker, trade policy is extremely
complex. Trade policy rarely comes in the form of a single, discrete act of
protection.14 Instead, trade policy is embedded in legislative acts made up of
complicated bundles of changes in the law regulating trade that even experts have
a hard time evaluating. NAFTA is a very interesting example. NAFTA itself is
Figure 5.5: Title VII petitions initiated.
14 Even in the days of the classic tariff system—say, 1870 to 1932—the political action revolved
around tariff acts with hundreds of line items. In addition, as research on 19th century voting suggests,
the social meaning of the tariff was highly variable across local electorates.
The Peculiar Political Economy of NAFTA 99
a document of over 300 pages (not including the national tariff schedules and
various other lists—with these the text runs over 2000 pages) covering trade in
goods, technical barriers to trade, government procurement, investment, services,
intellectual property, and the administrative and institutional conventions needed
for implementation; the agreement on environmental cooperation is another 20
pages; as is the agreement on labor cooperation; all of these need to be
implemented with specific national regulations in all three signatory countries.
Some of these regulations will affect trade in goods, others will affect trade in
factors of production; some will liberalize this trade, others will restrict it. In
addition, while NAFTA officially became law in 1994, as we saw above, full
implementation will not occur in sensitive sectors for periods of 10–25 years. In
this environment even trade economists might be expected to have fairly diffuse
priors with respect to NAFTA’s aggregate and distributional effects.15
Clearly, in the case of NAFTA, some form of “learning” had occurred, since
expressed evaluations had shifted fairly dramatically. What is completely unclear
is what triggered this reassessment. There was virtually no NAFTA-specific
information and the aggregate facts of the economy suggest the sort of
environment in which trade is usually viewed relatively benignly. This would
seem to be a virtually archetypal example of footloose preferences. Given the
weakened domestic political institutions supporting trade liberalization in the US,
and the concomitant likelihood of increased prominence for trade as a public
political issue, such footlooseness of trade preferences could play a significant role
in the future politics of trade.16 As proponents of liberal trading relations, it
behooves us to seek an understanding of such preferences.
While there may have been no change in the material or policy environment,
the period from 1991 to 1996 (and especially 1991–1993) was characterized by
an extraordinary amount of public discussion about NAFTA. It was the first
instance of highly public trade politics since the heyday of the classic tariff
system and, as such, it may tell us something about the politics of trade in years
to come. NAFTA was in the news, it was dinner table and cocktail party
conversation among non-economists, every politician had (and had to have) a
public opinion (sometimes different from their private opinion). Perhaps most
15 For the purposes of this chapter we abstract from the central importance to the US of locking in
Mexican economic reforms. This is consistent with nearly all of the political-economy research on free
trade areas in general, and NAFTA in particular. Though see the important series of papers by Ethier
(1998a–c, 1999, 2002).16 The events surrounding the WTO meeting in Seattle (30 November 1999), illustrate this point even
more clearly than does the politics of NAFTA. Nelson (1989) provides a discussion of the role and
transformation of US trade policy institutions for the support of trade liberalism. Nelson (1995)
discusses the collapse of those institutions in the early 1970s.
H.K. Hall and D.R. Nelson100
importantly, NAFTA was a major issue in the 1992 presidential election. Trade
economists approached minor celebrity status, and the unanimity with which
trade economists concluded that NAFTA was economically insignificant for the
US was truly stunning. With similar unanimity we heaped well-earned scorn on
dishonest claims about the “giant sucking sound”.17 But note that the timing is
important. NAFTA really takes off as public issue only when candidate Ross
Perot decided to focus on it, and this focus really comes after the change in
public evaluations. While both Patrick Buchanan and Perot were critical in 1991
and 1992, it is only in 1993 (especially with the publication of Perot’s book in
September and the Gore–Perot debate in November) that there was wide
discussion of the issue. Where economists appear to have been successful in
convincing Congress that NAFTA’s economic effects would be small, the public
was clearly confused and worried (with “undecided” responses rising rapidly in
1991 and 1992).
Clearly, high uncertainty and learning play a major role in the NAFTA case.
However, at least if we believe that the effects of NAFTA were likely to be small
(and probably positive), as we claimed at the time, this is a peculiar kind of
“learning”. That is, instead of converging on the “true” facts of the matter, the
public moved in the opposite direction. One useful way of approaching this
phenomenon is to consider it as an example of herd behavior. Herding occurs
whenever agents focus on a single behavior, with particular reference to cases in
which there are multiple plausible candidate behaviors. The phenomenon of herd
behavior is common enough that it has been used as the basis for a wide variety of
economic analyses based on such things as demand interdependence (Leibenstein,
1950; Schelling, 1978; Becker, 1991) and network externalities (Dybvig and
Spatt, 1983; David, 1985; Farrell and Saloner, 1985; Katz and Shapiro, 1985). An
alternative explanation in terms of information cascades has recently been
presented by Bikhchandani, Hirshleifer, and Welch (BHW, 1992), and further
developed in a number of later papers.18
In an environment where individuals can learn about the environment from both
private information and the behavior of others, an information cascade occurs
when agents ignore their private information and follow the information implied
by the behavior of others. Where the behavior of others is not perfectly
informative with respect to their private information, an information cascade
effectively traps socially useful information, thus permitting socially suboptimal
17 It is probably useful, given the recent vogue among trade economists for attacking free trade areas, to
recall that during the NAFTA debate support for NAFTA was seen as a litmus test of one’s status as a
serious economist.18 See Bikhchandani et al. (1998) and Nelson (1998) for surveys.
The Peculiar Political Economy of NAFTA 101
outcomes. Thus, the essential elements of the information cascade model are
coarse public signals and private signals of bounded accuracy. Without the first
assumption, the law of large numbers suggests that, with a sufficiently large
number of observations, the true state of the world is revealed (almost surely). In
most information cascade models this assumption takes the form that agents
observe the actions of other agents but not their signals, and that the actions are
imperfectly informative with respect to signals. Without the second assumption,
individuals might receive fully informative signals, allowing them to take actions
that would break the cascade. In addition, BHW (1992) make a number of
additional assumptions that permit a very simple expository model.
Consider the NAFTA case. Suppose that we start from an equilibrium and,
thus, from a cascade involving all citizens. Specifically, suppose we start from a
situation in which people have essentially no beliefs at all about NAFTA, and
possibly have a weakly held belief that trade liberalization (whether NAFTA or
multilateral) would have a negative effect on the economic interests of the US.19
That is, we assume that citizens as a whole believe that trade liberalization is
harmful, but this support of increased liberalization is highly conditional and
subject to large shifts. Now recall that in the run up to the fast-track vote,
virtually all respectable economists and the political leadership of both parties
argued very publicly that, essentially, NAFTA was no big deal economically for
the US, but that it was important politically (by being important economically
for Mexico). The result, as we have already noted, was strong public support for
NAFTA. However, the public had not learned that liberal trade was good, if by
“learned” we mean “identified the true state of the world”, but had simply shifted
to another weakly held prior. With the campaigns in 1992 and the debate over
NAFTA itself in 1993, the opposition forces began to receive considerable
greater visibility and, even though there was no change in the material
environment, people shifted their evaluations. Furthermore, once NAFTA was
passed by Congress, economists in general (and trade economists in particular)
not only lost interest in NAFTA, but began to argue that maybe NAFTA was not
such a great idea after all. It is not really relevant that the free trade
fundamentalist critique of NAFTA was that it distracted political attention from
19 Both assumptions here strike us as plausible approximations for the purposes of this example. What
poll data exist with respect to trade policy generally strongly suggests that a considerable majority
believes further liberalization will be harmful to the US economy—though there is little evidence of
support for general increases in protection (though support for sectoral increases is often strong)—see
Scheve and Slaughter (2001). Furthermore, there is very little evidence that citizens make much of a
distinction between preferential and multilateral liberalization. This makes sense. The difference in
complexity, from the point of view of a citizen, between NAFTA and, say, a GATT agreement, is
trivial.
H.K. Hall and D.R. Nelson102
broader trade liberalization. The public listens for the conclusion, not
the argument—especially when the argument is, at best, arcane.20 The result,
in the face of continued aggressive public relations against NAFTA, as we have
already noted, is that public opinion shifted back to opposition to NAFTA.
5.3. AN ILLUSTRATIVE MODEL
This section sets up a very simple political economy model of individual citizen
preferences for or against the NAFTA under uncertainty over the impact of the
agreement. Preferences over the trade agreement depend on two factors: an
economic effect; and a political effect. The latter reflects the claim that NAFTA
would help lock-in liberal political and economic reforms in Mexico. We will
assume that this effect is certain, but has a relatively small weight in citizen
welfare. The first reflects concerns with the economic effects of NAFTA on US
labor. Poll data seem to suggest that negative evaluations of NAFTA were
highly correlated with a perception of high labor adjustment costs in the US. If
Mexico is economically small, we suppose that there are no significant
adjustment costs; but if Mexico is large there will be significant adjustment
costs. Furthermore, if Mexico is large, the adjustment costs are taken to
outweigh the benefits of locking in Mexican reform. We will assume that these
preferences are common and reflected in a common evaluation function
VðE;PÞ U vðEðSÞÞ2 P; where E and P reflect the economic and political
factors, S denotes the economic size of Mexico, and vð·Þ is the value of
protecting displaced factors from adjustment costs, so a positive value of Vð·Þ
implies a preference for rejecting NAFTA.21
20 Abstracting from details of trade creation and trade diversion, which are characterized by complexity
considerations of the sort central to this chapter, it is notable that virtually all of the arguments against
regionalism, whatever their validity, are arguments that only an economist would love. Unlike the
simple models used to illustrate powerful, but difficult, notions of comparative advantage and gains
from trade, which are based on assumptions that isolate the key causal relation generating gains from
trade, the political economy arguments used to argue against free trade areas are based on assumptions
that seem unrelated to the core processes involved. Their purpose seems more to be to stiffen the spine
of the profession in its support of multilateralism than to persuade citizens or their representatives.21 That is we are assuming that citizens have a very simple form of sociotropic preferences (Kinder and
Kiewiet, 1981; Mutz and Mondak, 1997), i.e., they evaluate policies in terms of the way policy affects
community welfare. We simplify by assuming that the relevant community is the nation as a whole, and
that the evaluation of community welfare is unaffected by individual preference. These latter two
assumptions are generally false. We can prove the existence of cascades in more standard political
economy models, but the additional analytical freight does not produce any additional insight relative
to the central point of this chapter: footlooseness of policy preferences.
The Peculiar Political Economy of NAFTA 103
We now assume that Mexico’s size is uncertain. Each agent, therefore, faces
a decision under uncertainty of whether or not to reject the trade agreement
depending upon whether he/she expects it to be welfare improving. Specifically,
we will assume that each citizen observes a conditionally iid signal si ¼ {L; S}and that si . 0:5 if the true value of Mexico’s size is L and 12 si if the true value
is S: In addition to the private signal, citizens take their action in a known order
and each citizen observes the action (but not the signal) of all agents that precede
them in this order. Thus, after the first citizen decides whether or not to reject
NAFTA, all later deciders have two sources of information: one public, and one
private. It is common knowledge that all agents are Bayesian rational. If we adopt
BHW’s normalization of vðHÞ ¼ 1; vðLÞ ¼ 0; P ¼ 12; and letting gi be the
posterior probability that Mexico is large: E½vi� ¼ gi £ 1þ ð12 giÞ £ 0 ¼ gi; andVi . 0 if gi .
12: Finally, BHW adopt the tie-breaking convention that a citizen
that is indifferent between rejecting and accepting NAFTA chooses to accept.
This model is now identical to that of BHW’s “specific model” (BHW, 1992,
pp. 996–999).
The logic of cascades is straightforward in this model. Suppose that the
first citizen observes s1 ¼ L: She has no additional information, so V1 ¼
12 12. 0 and she rejects NAFTA. For citizen 2 there are two possibilities:
s2 ¼ L and citizen 2 rejects; or s2 ¼ S; so citizen 2 computes E½v2� ¼12; V2 ¼
0; and by the tie-breaking rule accepts. Citizen 3 is in one of three cases: both
the previous citizens reject NAFTA, so citizen 3 rejects independently of her
private signal; both previous citizens accepted, so citizen 3 accepts
independently of her private signal; or 1 and 2 split, in which case 3 is in
the same situation as 1—i.e., her private signal determines her action. Either
of the first two cases is a cascade: all later citizens ignore their private signals
and adopt the same action of those preceding them. In this simple framework
it is easy to see that: cascades occur with probability 1; cascades occur more
quickly the further is s from 0.5; and good cascades occur more often the
further is s from 0.5. With reasonably high uncertainty, s not terribly far
from 0.5 (as we have argued may have been the case with NAFTA), but with
a large number of citizens, everyone knows that they are in a cascade almost
surely. This is the opening wedge for BHW’s other key result that: (1) once
an informational cascade has begun, individuals still value public information
and (2) a small amount of public information can reverse a cascade. That is,
even a strong informational cascade rejecting NAFTA can be suddenly
reversed into an equally strong cascade supporting the agreement. The key
result here is their result (3) “the release of a small amount of public
information can shatter a long-lasting cascade, where a ‘small amount’ refers
H.K. Hall and D.R. Nelson104
to a signal less informative than the private signal of a single individual”
(BHW, 1992, p. 1005).22
5.4. CONCLUSION: ON ECONOMISTS AS PARTICIPANTS
IN THE POLITICS OF TRADE POLICY
It is important to be clear that, while the class of models considered here is
positive, they are not predictive over the domain of final political-economic
outcomes. These models do not avoid, in fact they rest on (or, more accurately,
provide a formal representation of) the fact that, with appropriately chosen
priors, we can reproduce virtually any final outcome.23 Thus, any predictions of
these models with respect to final outcomes are vacuous. Nonetheless, we hope
that this chapter suggests that ignorance and learning in a social context are
issues of first-rate importance, both as empirical phenomena and as potential
determinants of trade policy outcomes. Furthermore, there are, we think, several
implications of these models (and of ignorance/learning more generally) for us
as policy analysts. We consider two: one with respect to evaluating the
predictive content of our positive models; and the second with respect to the role
of economists in the public discourse over trade policy.
There is no substitute for basing predictive political-economy models on
political and economic fundamentals. We have good reason to expect such
fundamentals to play a central role in determining trade policy, andwe have equally
good reason to predict the direction of the effects. However, if learning effects also
play a, largely unpredictable, role, we also need to expect prediction errors that are
occasionally large. That is, the right kind of ignorance can yield wildly different
outcomes from those predicted by models. An excellent example, away from the
trade policy focus of this chapter, is the poor performance of macro political-
economy models in the 1992 Bush–Clinton election (Haynes and Stone, 1994).
While the data seem to show recovery of the economy, and thus success for the
22 BHW (1992) develop their analysis in a somewhat more general informational environment
involving a sequence of possible signals and a more careful formalism. More importantly, it would be
straightforward to extend their analysis to the case of preference heterogeneity, as generated, for
example, in a standard endogenous tariff model, under full information about citizen types. In that case,
while actions (e.g., accept, reject, abstain) will vary, so there need not be complete herding, there will
still be good and bad cascades with strictly positive probability. More interesting possibilities emerge
with type heterogeneity and uncertainty about types. Smith and Sørensen (2000) develop this case in
detail. This is an important topic for future research, but beyond the needs of the simple point made in
this chapter.23 It is the reproduction of outcomes that is most worrying in social scientific analysis. Thin predictions
(unless strictly unfalsifiable) will not last, but compelling post-dictions (“stylized facts”) can sustain
empirically weak theoretical analyses for long periods.
The Peculiar Political Economy of NAFTA 105
incumbent, there was widespread perception that the economy was still in an
economic crisis (“It’s the economy stupid”). The result, as they say, is history.
As noted in the introduction and suggested by the discussion of the NAFTA case,
one of the most interesting implications of learning models with information
cascades as a prediction is the suggestion of a major role for policy analysts. Some
of the recent attempts to justify an active policy role for economists turn on difficult
philosophical issues of freedom of choice that seem rather removed from the actual
practice of participation by economists in the public policy discourse. The advice
on how and whom to advise that emerges from this kind of argument seems of
limited use.24 The problem seems to emerge from taking our models seriously
where we should not. We have just argued that these models serve a very useful
positive purpose in understanding and predicting public policy. However, when we
abstract from complexity and uncertainty in the interest of building parsimonious
models, we have abstracted from the most obvious warrant for an active advisory
role. When agents, whether citizens or policy-makers, are highly uncertain about
the workings of the economy (i.e., most of the time), expert advice can have a
substantial effect on final outcomes via precisely the channels identified in learning
models.25 It makes perfectly good sense for citizens and politicians to listen to, and
even to seek out, the advice of economists because that advice is better informed
thanmuch of the policy advice that is given during a political process—though note
that this need not be even vaguely perfect information. A public signal of strong
agreement among economists, especially when supported by compelling evidence,
during a political process can have the effect of a public information release in the
models discussed above. Even if policy-makers, or citizens, believe that this
information is less informative than any individual privately observed signal, such a
public release can have the effect of reopening the public discussion and
dramatically shifting the structure of governmental or public opinion.26
24 This is the entering wedge for Dixit’s (1997) comment on O’Flaherty and Bhagwati (1997).25 Just as trade and political economy models abstract from informational issues to focus on the causal
forces of most immediate interest, learning models of the sort developed in this chapter abstract from a
variety of complexities to highlight the effects of ignorance and learning in a social context. In
particular, these models abstract from the important, and complementary, forces that make advisors
participants (in a game theoretic sense) in the political process. Dixit (1997) provides a very nice
sketch, with appropriate references, of models which highlight the role of advice giving in a strategic
environment with asymmetric information.26 Although we are presumably better informed, as we noted above, economists are at least as prone to
being trapped in cascades as any other rational agents engaged in learning about the world, possibly
more so. Because economists have very similar understandings of the workings of the economy—due
to strong socialization—we may behave more like the agents in the BHWmodels than any other group.
Smith and Sørensen (2000) find much more complex aggregate behavior characterizing groups with
heterogeneous preferences than in groups with homogeneous preferences.
H.K. Hall and D.R. Nelson106
It seems to us that a focus on this non-strategic, informational role of economic
advice has useful implications. To the extent that the warrant for advice-giving is
uncertainty, as much about the working of the economy as about simple facts, it
seems particularly fruitless to give advice based on the presumption that those
receiving it are well-trained economists.27We need to convince our auditors that a
consensus on fundamental issues related to, say, trade policy, exists, and we need
to do so in ways that are clear to relatively engaged, relatively intelligent
non-economists. This clearly means that complicated arguments, requiring many
closely argued steps, and knowledge of economic theory, are likely to be
unsuccessful. However, as Matthew Slaughter (1999) argues, while outright lies
may be successful in the short-run (e.g., “NAFTA will create thousands of jobs”),
sooner-or-later they are likely to backfire. The most successful of our public
representatives—e.g., Milton Friedman, Alan Blinder, Paul Krugman—seem to
identify simple but compelling metaphors, which are mixed with a small number
of striking facts, to argue for a single clear policy point. Finally, it should be noted
that, if social learning does not produce knowledge of the intertemporally sturdy
type, but rather of the type suggested by informational cascade models, we need to
be prepared to stay engaged in the public discourse beyond the passage of any
particular piece of legislation.
ACKNOWLEDGEMENTS
The authors would like to thank participants at the 1999 ASSA meetings, New
York, NY and the 25th Anniversary Meeting of the International Economic Study
Group, September 9–11, 2000, Isle of Thorns, Sussex. In particular, we are
grateful to comments from Alan Winters and Devashish Mitra.
REFERENCES
Bailey, M. and Brady, D. (1998). Heterogeniety and representation: the senate and free
trade. American Journal of Political Science, 42, 524–544.
Baldwin, R. and Magee, C. (2000). Is trade policy for sale? Congressional voting on recent
trade bills. Public Choice, 105, 79–101.
Becker, G. (1991). A note on restaurant pricing and other examples of social influence on
price. Journal of Political Economy, 99, 1109–1116.
27 While admirable, the attempt to transform citizens and policy-makers into economists is almost
certainly doomed to failure.
The Peculiar Political Economy of NAFTA 107
Bikhchandani, S., Hirshleifer, D. andWelch, I. (1992). A theory of fads, fashion, custom, and
cultural change as informational cascades. Journal of Political Economy, 100, 992–1026.
Bikhchandani, S., Hirshleifer, D. and Welch, I. (1998). Learning from the behavior of
others: conformity, fads, and informational cascades. Journal of Economic Perspectives,
12, 151–170.
Boadu, F. and Thompson, M. (1993). The political economy of the U.S.–Mexico free trade
agreement: analysis of the congressional fast track vote. Journal of Agricultural and
Applied Economics, 25, 27–35.
Conybeare, J. and Zinkula, M. (1994). Who voted against the NAFTA.World Economy, 19,
1–12.
David, P. (1985). Clio and the economics of QWERTY. American Economic Review. 75,
332–337.
Destler, I. M. (1995). American Trade Politics, 3rd edn., Washington, DC: Institute for
International Economics.
Dixit, A. (1997). Economists as advisers to politicians and society. Economics and Politics,
9, 225–230.
Dybvig, P. and Spatt, C. (1983). Adoption externalities as public goods. Journal of Public
Economics, 20, 231–247.
Ethier, W. (1998). The international commercial system. Essays in International Finance,
#210.
Ethier, W. (1998). The new regionalism. Economic Journal, 108, 1149–1161.
Ethier, W. (1998). Regionalism in a multilateral world. Journal of Political Economy, 106,
1214–1245.
Ethier, W. (1999). “Multilateral roads to regionalism,” in International Trade Policy and
the Pacific Rim, J. Piggot and A. Woodland (eds.), London: Macmillan, 131–152.
Ethier, W. (2002). Unilateralism in a multilateral world. Economic Journal, 112, 266–292.
Farell, J. and Saloner, C. (1985). Standardization, compatibility, and innovation. Rand
Journal of Economics, 16, 70–83.
Haynes, S. and Stone, J. (1994). Why did economic models falsely predict a Bush landslide
in 1992. Contemporary Economic Policy, 12, 123–130.
Holian, D., Krebs, T. and Walsh, M. (1997). Constituency opinion, Ross Perot, and roll-call
behavior in the U.S. house: the case of the NAFTA. Legislative Studies Quarterly, 22,
369–392.
Kahane, L. (1996). Congressional voting patterns on NAFTA: an empirical analysis.
American Journal of Economics and Sociology, 55, 395–409.
Kahane, L. (1996). Senate voting patterns on the 1991 extension of the fast track trade
procedures: prelude to NAFTA. Public Choice, 87, 35–53.
Kamdar, N. and Gonzalez, J. (1998). An empirical analysis of the U.S. senate vote on
NAFTA and GATT. International Advances in Economic Research, 4, 105–114.
Katz, M. and Shapiro, C. (1985). Network externalities, competition and compatibility.
American Economic Review, 75, 424–440.
Kinder, D. and Kiewiet, D. (1981). Sociotropic politics: the American case. British Journal
of Politics, 11, 129–161.
H.K. Hall and D.R. Nelson108
Leibenstein, H. (1950). Bandwagon, Snob and Veblen effects in the theory of consumer
demand. Quarterly Journal of Economics, 64, 183–207.
Levy, P. (1998). “Learning from trade agreements,” MS, Department of Economics, Yale
University.
Lustig, N. (1998). Mexico: The Remaking of an Economy, 2nd edn., Washington, DC:
Brookings.
Mayer, F. (1998). The Science and Art of Policy Analysis, New York: Columbia University
Press.
Mutz, D. and Mondak, J. (1997). Dimensions of sociotropic behavior: group-based
judgements of fairness and well-being. American Journal of Political Science, 41,
284–308.
Nelson, D. (1989). The domestic political preconditions of US trade policy: liberal structure
and protectionist dynamics. Journal of Public Policy, 9, 83–108.
Nelson, D. (1995). “Political economy of protection for the US automobile industry,” in
Political Economy of American Trade Policy, A. Krueger (ed.), Chicago: University of
Chicago Press, 133–191.
Nelson, D. (1998). “Footloose policy preferences and trade policy: the role of non-strategic
social learning and information cascades,” MS, Murphy Institute of Political Economy.
O’Flaherty, B. and Bhagwati, J. (1997). Will free trade with political science put normative
economists out of work? Economics and Politics, 9, 207–219.
Pastor, M. and Wise, C. (1994). The origins and sustainability of Mexico’s free trade
policy. International Organization, 48, 459–489.
Schelling, T. (1978). Micromotives and Macrobehavior, New York: Norton.
Scheve, K. and Slaughter, M. (2001). Globalization and the Perceptions of American
Workers, Washington, DC: Institute for International Economics.
Slaughter, M. (1999). Globalisation and wages: a tale of two perspectives.World Economy,
22, 609–629.
Smith, L. and Sørensen, P. (2000). Pathological outcomes of observational learning.
Econometrica, 68, 371–398.
Steagall, J. and Jennings, K. (1996). Unions, PAC contributions, and the NAFTA vote.
Journal of Labor Research, 17, 515–521.
Steagall, J. and Jennings, K. (1996). Unions and NAFTAs legislative passage:
confrontation and cover. Labor Studies Journal, 21, 61–79.
Ten Kate, A. (1992). Trade liberalization and economic stabilization in Mexico: lessons of
experience. World Development, 20, 659–672.
Thorbecke, W. (1997). Explaining house voting on the north American free trade
agreement. Public Choice, 92, 231–242.
Tornell, A. and Esquivel, G. (1997). “The political economy of Mexico’s entry into
NAFTA,” in Regionalism versus Multilateral Trade Arrangements, T. Ito and A.
Krueger (eds.), Chicago: University of Chicago Press, 25–55.
Weiss, J. (1992). Trade liberalization in mexico in the 1980s: concepts, measures and short-
run effects. Weltwirtschaftliches Archives, 128, 711–726.
The Peculiar Political Economy of NAFTA 109
CHAPTER 6
Interest and Ideology in the 1988
Omnibus Trade Act: A Bayesian
Multivariate Probit Analysis
ALOK K. BOHARA,a KISHORE GAWANDE,b,*
aDepartment of Economics, MSC05 3060, 1, University of New Mexico,
Albuquerque, NM 87131-0001, USAbGeorge Bush School of Government and Public Service, Texas A&M University,
4220 TAMU, College Station, TX 77843-4220, USA
Abstract
The landmark 1988 Omnibus Trade and Competitiveness Act marked an
important change in US trade policy. Using new Bayesian methodology that
facilitates model comparisons in a multivariate probit model, we study whether
ideology or interest determined congressional voting. We find that Senate voting
was driven equally by economic factors, mainly constituent interests, and political
factors, mainly party. House voting, on the other hand is largely determined by
party and ideology. We also find a common underlying factor not accounted by
the set of explanatory variables. We call this factor “trade ideology”, as distinct
from conventional liberal-conservative ideology.
Keywords: Multivariate probit, congressional voting, interest, ideology
JEL classifications: D72, C35, F13
*Corresponding author.
E-mail address: [email protected]
…I can note that the Administration will resist the following: First that
we would impose a general surcharge…Second, we oppose sector-
specific protection, such as establishing import quotas for individual
producers…Third, we oppose mandatory retaliation.And, finally, Mr.
Chairman, it comes as no surprise, I know, that we will oppose limits on
Presidential discretion.
Treasury Secretary James Baker, at Senate hearings before the Finance
Committee, 100th Congress, first session, February 19, 1987.
I see a dramatic change in the policy of this Administration, much more
aggressive trade policy and tougher trade policy……The fact is that trade
is the handmaiden of all other policy considerations of the U.S.
Government and it will continue to be so unless we have some limits on
the President’s discretion.…We need a trade policy that our trading
partners can predict and I maintain that requires limits on the President’s
discretion not to act.
Chairman of the Finance Committee, Senator Lloyd Bentsen, at Senate hearings
before the Finance Committee, 100th Congress, first session, April 2, 1987.
In considering any amendment to Section 301, I hope youwill ask one key
question: Will this amendment help or hurt the ability of U.S. negotiators
to pry open a foreign market to U.S. exports? It really comes down to this
issue: Is Section 301 an import-relief law, or is it a negotiating
tool?……What you have to have is a credible threat of retaliation…
US Trade Representative General Counsel Alan Holmer, at Senate hearings
before the Finance Committee, 100th Congress, first session, March 17, 1987.
6.1. INTRODUCTION
The literature on the determinants of voting behavior has a long and rich history in
political science and economics. Vigorous debate persists about the importance of
“ideology” versus “interest” in explaining voting behavior (Kau and Rubin, 1979;
Kalt and Zupan, 1984; Peltzman, 1984). Adherents of ideology (Poole and
Rosenthal, 1997; Kalt and Zupan, 1984) emphasize party and liberal-conservative
ideology as the main determinant of politicians’ voting behavior, while
proponents of interest emphasize lobbying and constituency interests as
A.K. Bohara and K. Gawande114
fundamental determinants of congressional voting (Kau et al., 1982; Peltzman,
1984, 1985; Irwin, 1994). The recent literature presumes that the two chambers
respond fundamentally differently. Bailey and Brady (1998) focus on Senate
voting and constituency interests. Levitt and Snyder (1995) show that in the House
parties influence policy outcomes that favorably redistribute income to their
districts. Adler and Lapinski (1997) show that since committees are composed of
members representing districts that have a demonstrated need for policy benefits
under the jurisdiction of those committees, constituency interests matter to House
voting. We visit the ideology versus interest debate in the setting of a landmark
legislation, the 1988 Omnibus Trade and Competitiveness Act.
The Act changed the nature of US trade policy from its previously free-trade
orientation. Although the change in US trade policy had been in evidence since
the 1979 Trade Act, the 1988 Act cemented the change. It made US trade policy
more aggressive and activist, and provided a basis for more strategic trade
policy. It was enacted during the term of a President strongly in favor of free
trade, and hence required forceful legislation in Congress. It lessened the ability
of the President to make discretionary decisions without consulting Congress, a
factor that contributed to previous free-trade stance of the US Enacted during a
time of weak world institutions in trade, the Super 301 provisions of the 1988
Act allowed greater unilateralism. This has disturbed political economists from
the free-trade persuasion (e.g., Bhagwati and Patrick, 1990) while giving
legitimacy to the activist strategic school (e.g., Krugman, 1986). Roll call votes
taken during legislation of the 1988 Act spanned a range of issues, providing for
a rich setting for examining the relative importance of interest versus ideology.
Tosini and Tower (1987) were among the first to study congressional voting on
trade bills. They considered voting on the protectionist 1985 Textile Bill, focusing
on variables that measured economic interest behind voting. Their study found
that senators and representatives responded to a variety of influences while voting
on the bill, including campaign contributions, union power, unemployment in
their constituencies, and the economic importance of the textile industry in their
constituencies. Many of the variables used by Tosini and Tower have informed the
economic variables used in this study.
We methodologically extend the study of roll call voting data by employing a
multivariate probit model (MVP). Any analysis of roll call votes on connected
issues, many occurring within days of each other, must come to grips with the fact
that there will be jointness across the votes. The correct methodology is, therefore,
a multivariate approach. That is, each congressman’s votes on a set of bills are
correctly modeled as a draw from a multivariate distribution. Merely conditioning
on a set of exogenous variables in a single equation analysis misses crucial
information present in the error correlations, leading to inefficient estimation.
Interest and Ideology in the 1988 Omnibus Trade Act 115
Our finding of statistical significance and large magnitudes on the error
correlations bears this out.
Methodologically, the chapter showcases an important new statistical
technology due to Chib and Greenberg (1998), namely the Markov Chain
Monte Carlo (MCMC) method. This method is useful whenever high-dimensional
integration is required in estimation, which is true of Bayesian analysis in general
(and classical analysis of the MVP). With a reasonably high dimension, say,
greater than two, the estimation of the MVP model by traditional methods is
cumbersome and becomes increasingly inaccurate as the number of dimensions
increases. The MCMC method also allows model comparisons with ease. Non-
nested model comparisons of models emphasizing ideological factors versus
models emphasizing interest are featured here. We believe it to be an innovation in
the legislative studies literature.
In Section 6.2 the historical context of the Omnibus bill is described. It is
followed by a description of the methodology. The set of hypotheses central to the
empirical work and the data used to examine them are then described in Sections
6.3–6.5. The empirical results are analyzed in Sections 6.6 and 6.7. Section 6.8
concludes via conventional significance tests as well as formal model comparisons.
6.2. BACKGROUND: THE TRADE OMNIBUS ACT
Table 6.1 presents a summary of the features of the Omnibus Act and the route the
bill took to become law. Table 6.2 details the roll calls. The genesis of the 1988
Act lay in a strategic and relatively protectionist bill that was passed by a large
majority in the House in 1986 (HR 4800). However, the Senate refused to consider
it since support for the bill fell far short of the two-thirds majority required to
override a certain presidential veto. The 99th Congress saw a shift in power in the
Senate, with 54 Democrats and 46 Republicans, and the stage was set for activist
trade legislation. The House Ways and Means Committee, with primary
jurisdiction over trade matters, authored bill HR3 in 1987 under the shrewd
chairmanship of Dan Rostenkowski (D-IL). It was marked up by four other
committees making for an Omnibus bill.1 The most controversial part of HR3 was
the Gephardt amendment, named after Rep. Gephardt (D-MO), who would later in
the year run for the Presidential nomination. The Gephardt amendment would
make it mandatory for the President to retaliate against partners with “excess”
trade surpluses vis-a-vis the US (i.e., if their exports totaled 175% of imports)
1 Energy and Commerce added strategic amendments to promote exports in high-tech US goods;
Foreign Affairs added amendments furthering US exports, and on relaxing export controls; banking
added strategic amendments on exchange rates and third world debt.
A.K. Bohara and K. Gawande116
Table 6.1: Evolution of the Omnibus bill: house, senate, conference, and final versions.
Features of the House bill Features of the Senate bill
H.R.3
(Introduced
in Ways
and Means)
(1) Presidential negotiating authority: approval
of fast track for 5 years
(2) Transfer of authority over Section 301
(unfair trade practices) cases and retaliation to
USTR—a lower level of official making this
decisions rather than the president. Also, authority
over Section 201 (escape clause), and 337
(intellectual property rights)
transfers from president to USTR
(3) Section 301: mandatory action against agreement
violation but otherwise up to president’s discretion.
“Special 301” allowed self-initiation on an annual
basis of 301 cases against priority foreign
countries who failed to protect intellectual
property rights
(4) Gephardt bill: USTR required to retaliate
against unfair trade practices; mandatory
action against foreign countries to reduce
their bilateral trade imbalance by 10%
per year until 1991
(5) AD/CVD: extended to high-tech and
agricultural goods from its traditional focus
on steel and other heavy industries
(6) Bryant amendment: disclosure
of foreign ownership in a US
business to be reported to SEC
S1420 (Introduced
in Finance)
(1) Presidential negotiating authority:
possibility of withdrawal by Congress
of president’s access to fast track
(2) Section 201: shift of earlier focus
on protection to prevent injury to
protection to assist industry to adjust
to import competition; mandatory
funding of such adjustment. Less
ability for president to deny ITC
decision for protection; but set higher
standard for industry in order to file
under 201. Expansion of Trade
Adjustment Assistance
(3) Section 301: severely reduced
president’s discretion not to act in
301 cases. Expansion of scope of
practices actionable under 301. Amend
Section 301 to include “mandatory but
not compulsory” retaliation against
unjustifiable or discriminatory foreign
(e.g., Japan) practices
(4) Other issues: telecommunications
trade opportunities; intellectual
property rights protection; access
to foreign technologies; auctioned quotas
(continued)
InterestandIdeologyin
the1988OmnibusTradeAct
117
Table 6.1: (continued)
Features of the House bill Features of the Senate bill
Michel (R-IL) substitute for H.R. 3 (#78)
passed on 4/30/87, 290–137. It excluded
Gephardt (but adopted Rostenkowski’s
alternative which was similar to Gephardt
except for “surplus reduction” measures),
Bryant, and “Buy America”
provision. Else same as above
S1420 passed (7/21/87), 71-27 (#208) with
addition of 34 (out of 160 proposed)
amendments. “Super 301” introduced:
relation to target consistently protectionist
countries. Amended S1420 close
to original version above
Features of the Conference Bill
Rostenkowski (Chair, House Ways and Means) started by proposing to drop most protectionist sections, and Bentsen (Chair, Senate Finance)
reciprocated the sentiment. The bill adopted in conference, still called H.R.3 was a less protectionist version of the above bills, but contained their
activist and strategic elements. It dropped Gephardt but included a labor-backed plant closing provision which required employers to give 60 days
notice of plant closings or layoffs. It passed 312–107 on 4/21/88 in the House (#66), and 63–36 on 4/27/88 in the Senate (#110), and was sent to
President Reagan
Final Omnibus bill
The bill was vetoed by President Reagan on 5/24/8 8, citing the plant closing provision. The House vote on the same day (#150) overrode the veto, but
the Senate vote of 61–37 (#169) on 6/8/88 failed to override. A new bill in the House, H.R. 4848, close to the vetoed bill but without the plant closing
provision, was voted in 376–45 by the House (#231) on 7/13/88, and 85–11 (#288) by the Senate on 8/3/88. It was signed into law on 7/23/88. The
plant closing provision made its way through the Senate and House as a separate bill (S2527) and became law in August 1988
Abbreviations: USTR, US Trade representative; AD/CVD, Antidumping Duty/Countervailing Duty; TAA, Trade Adjustment Assistance; SEC,
Securities and Exchange Commission. Sections 201 and 301 refer to same Sections of the Trade Act of 1974.
A.K.Bohara
andK.Gawande
118
if this excess was due to unfair trade practices.2The amendment passed with amere
four-vote margin on the House floor, but would be dropped in conference. It was
anticipated that Reagan would veto the bill since, other than being an activist bill
even without the Gephardt amendment, it sought to curb discretion of the executive
in granting or denying protection. That discretion had been used by Reagan to
reverse the decision of the International Trade Commission (ITC), the independent
bipartisan body that made determinations under Sections 201, 301 and 337 of the
US Trade Act. The ITC had found for the petitioners seeking protection.
In the Senate, the Finance Committee with primary jurisdiction over trade
matters, led by Sen. Bentsen (D-TX), proposed a Senate bill similar in content to
HR3. While it was an activist bill, it was less protectionist. It contained no
Gephardt-like provisions. However, it contained a labor-backed provision on plant
closing (requiring employers in large firms to give 60 days prior notice to any
plant closings or layoffs) that would be the basis for Reagan’s veto of the Omnibus
bill. The Senate failed to override the veto, and this provision was dropped to
ensure passage of the 1988 Omnibus Trade and Competitiveness Act. The plant
closing provision succeeded separately and became law in 1988. Though the
Omnibus bill was an activist “fair trade” bill, its authors Rostenkowski and
Bentsen pushed to drop explicitly protectionist provisions from the bill. They
foresaw the need for substantial support from moderate Republicans. Hence, the
protectionist Textile bill was not packaged into the Omnibus bill, Gephardt’s
provision was dropped, as was the plant closings provision after the veto.3
6.3. ANALYSIS OF THE MULTIVARIATE PROBIT MODEL
The econometric methodology employed in this chapter is described here at an
intuitive level. For full technical details we refer the reader to Chib and
Greenberg (1998) and Chib (1995). The Bayesian method of estimating the MVP,
due to Chib and Greenberg (1998), is employed in this chapter because it is
appropriate for the study of multiple roll call voting on related issues (Table 6.2).
The method facilitates formal model comparisons, which is a major focus of this
study. While it allows the incorporation of prior information formally into the
estimation, in this study we use only weak priors, which are easily dominated
2 The genesis to the Gephardt amendment lay in an import surcharge bill during 1986 authored by
Rostenkowski, Gephardt and Senator Bentsen, all key players in the Omnibus bill. But later
Rostenkowski and Bentsen distanced themselves from it and it became Gephardt’s amendment.3 There may have been implicit agreements among the architects of the Omnibus bill and these other
bills about reciprocal support. To maintain focus, we do not formally test for logrolling here, leaving
this open for future work.
Interest and Ideology in the 1988 Omnibus Trade Act 119
Table 6.2: Roll call votes analyzed in the multivariate probit models.
House bills Senate bills
Bill # Short description Date Vote Bill # Short description Date Vote
72 Gephardt(D-MO). Mandatory retaliation
to lower trade surplus
4/29/87 218–214 158 Bentsen (D-TX). Against
Specter’s motion for private
right of action to sue
against unfair trade
6/25/87 69–27
77 Michel (R-IL). Drop Gephardt, Wylie,
and Bryant (see Table 6.4) amendments
4/29/87 156–268 175 Bradley (D-NJ)/Packwood(R-OR).
Against requiring President to
keep oil imports below 50%
of consumption
7/1/87 55–41
78 Omnibus Bill, House version. Like
Michel’s version (77). Activist,
Strategic. See Table 6.4
4/30/87 290–137 178 Packwood (R-OR). Allow
President discretion over
ITC cases on account of
national interest
7/7/87 41–55
426 Michel’s (R-IL) motion to instruct
House conferees to concede
Gephardt amendment
11/10/87 175–239 179 Domenici (R-NM).
Expand definition of
unfair trade practices
7/8/87 71–28
66 Omnibus Bill passage, Conference version.
Gephardt dropped. Plant closing included
4/21/88 312–107 208 Omnibus Bill passage,
Senate version. See Table 6.4
7/21/87 71–27
231 Final Omnibus Bill Passage: After
dropping plant closing and Gephardt
7/13/88 376–45 110 Omnibus Bill passage, Conference
version. See Table 6.4
4/27/88 63–36
285 Bensten (D-TX). Defend the
provision in the Bill to transfer
authority from President to USTR
8/3/88 69–26
Notes: full descriptions of roll calls available from authors. Other roll calls taken on the Omnibus bill but not included in the analysis: #150(S),
#77(H), #206(S), #169(S), #114(H), #197(H), #319(H) #132(S), #250(S), #190(S), #229(H) #267(H), #341(H), #426 (1988) (H), #225 (S), 288 (S),
329 (S), 332 (S).
A.K.Bohara
andK.Gawande
120
by data evidence. Hence, for all practical purposes the results are based on the
data, not prior information, so as to service the widest readership. Weak priors
also adequately represent our prior beliefs about the determinants of voting on the
Omnibus Act.
Consider the J-variate probit model where Yij is a random variable denoting a
binary 0=1 response on the ith observation on the jth variate. Let yi ¼ ðyi1;…;yiJÞ0;
1 # i # n; be the J-variate realization on the ith observation. The MVP model is
conditioned on the parameters b : k £ 1; and S : k £ k and a set of explanatory
variables Xi; as follows:
yi ¼ Xibþ ei; ð1Þ
where the data yi : j £ 1 as described above, Xi is a ðJ £ kÞ matrix of explanatory
variables given as diagðxi10;…; xiJ
0Þ; where xij : kj £ 1 is the set of kj explanatory
variables in the jth equation (hence k ¼P
Jj¼1 kjÞ; b : k £ 1 is the vectors of
coefficients on the explanatory variables, and ei : J £ 1 is the (correlated) vector
of residuals with zero mean and correlation matrix S : J £ J: In Bayesian analysisof the model, b and S are assigned a prior distribution (here these prior
distributions represent ignorance, so inferences are from the data, not priors).
Inference proceeds with the analysis of the posterior distribution, which combines
the prior probability density function with the data likelihood.
The estimation of the posterior mean of the parameters (used to make
inferences) and the posterior ordinate at the posterior means (used to perform
model comparisons) are performed via a powerful computational method
involving MCMC sampling. Estimation by MCMC methods has been used
effectively in the Bayesian analysis of several econometric models (e.g., Chib and
Greenberg, 1996). In applications of MCMC methods an overriding feature is
that the joint density (“posterior density” in Bayesian applications, or “data
likelihood” in non-Bayesian applications) is not analytically well defined. But it
can be broken down into the product of conditional densities that are analytically
well defined, from which synthetic samples may be taken. The Gibbs sampler
derives its utility from the fact that generating sequences of samples from the
conditional distributions upon updating the conditioning parameters, results in a
sample that is actually generated from the joint density itself. That sample can be
used to compute the posterior mean or maximum likelihood estimate, and also
the ordinate of the posterior density or value of the likelihood function. Chib and
Greenberg (1998) provide explicit conditional distributions and sampling methods
to generate samples from the joint multivariate probit density, from which b and Sare estimated. Using Chib’s (1995) method the sample is also used to perform
formal comparisons of competing models of legislative behavior.
Interest and Ideology in the 1988 Omnibus Trade Act 121
6.4. THEORY AND MEASUREMENT
This study is concerned with the empirical investigation of three issues in
legislative behavior:
† determinants of voting behavior on redistributive issues, specifically trade
legislation,
† relative importance of ideology and interest, and
† whether and why determinants of voting may vary across chambers.
Several studies we reference below, which provide the basis for the explanatory
variables in our econometric model, present illuminating though diverse answers
to the first question. However, most empirical studies of voting are content with
testing hypotheses against no specific alternatives, and are silent on the second
issue. We are aware of no studies that formally weigh their own hypotheses
against well-specified alternatives, which is a distinguishing feature of this study.
On the third question, most studies to date are again silent. We do not pretend to
provide a full, or even adequate, answer to this third question. Since the analysis of
a closely related set of votes across both chambers is possible, we are presented
with a natural experiment that allows us to shed light on possible inter-chamber
heterogeneity. Our hope is that theoretical explanations able to formalize what we
find empirically will be forthcoming. The remainder of this section is devoted to
describing our empirical models, which are founded on a set of recent theoretical
and empirical studies. Baldwin (1986) and Conybeare (1991) model the electoral
market for protection as the equilibrium outcome of demand and supply factors. In
this model, existing protection leads to a demand for more protection. Since
protection has income effects (more protection raises incomes), and is a normal
good, districts/states with greater protection will desire more of it. On the supply
side, a politician is likely to vote for greater protection the lower is her opportunity
cost of doing so. A politician is also likely to vote for greater protection depending
on her institutional capacity to supply protection. In our setting Democrats
possessed the larger institutional capacity to vote for trade activism.
In the present setting, a reduced form equation predicting voting on, say, issues
that are for activism or protectionism, we would expect to find a positive sign on
previous protection, a positive sign on variables decreasing in the opportunity cost
of providing protection, and negative sign on party (the variable PARTY being
measured as a binary variable: 1 if Republican, and 0 if Democrat). To measure
existing protection we employ a tariff measure (TARIFF), and a nontariff barrier
measure (NTB). To measure the opportunity cost of providing protection we
employ the net-exports-to-shipment-ratio (NETEXP) and the unemployment rate
A.K. Bohara and K. Gawande122
(UNEMP). Low values of NETEXP indicate comparative disadvantage and hence
lower opportunity cost of voting for protection, while higher values of UNEMP (in
a time of economic boom) indicate industries in distress and hence lower
opportunity cost of voting for protection. NETEXP is thus expected to have a
negative (positive) sign on votes for protection (free trade). On votes for
protection, a positive sign is expected on UNEMP.
Irwin’s (1994) study of British voting on free trade in 1906 demonstrates the
importance of sectoral employment in the pattern of voting. We similarly measure
constituency interests by employment shares in food processing (FOOD), which
cross-industry studies of protection have shown to be a pro-protectionist sector.4
Bailey and Brady (1998) and Bailey (2001) theorize that the relationship
between constituency characteristics and roll call voting varies with party. This
hypothesis about inter-party heterogeneity is based on the premise that there are
systematic differences across parties in how electoral coalitions are built. They
find evidence in favor of inter-party heterogeneity from 1993–1994 Senate voting
on key issues related to NAFTA and GATT. We incorporate the Bailey–Brady
hypothesis by interacting TARIFF, NTB, and NETEXP with PARTY, where
PARTY equals 1 for Republicans and 0 for Democrats.5 Significance of the
interaction terms are evidence of heterogeneity in how a politician’s party
qualifies her response to constituency interests.
Irwin and Kroszner (1999) study the Republican conversion to trade
liberalization during the 1940s after the Republican enactment of the Smoot–
Hawley tariffs of the 1930s. Their theory favors interest over ideology. Interacting
party with exports, they find that sensitivity to exporting industries in their
constituencies, rather than any shift in ideology, was the main factor responsible
for the Republican “conversion” to the Democratic free-trade initiative.
Empirically, their finding reinforces Baily and Brady’s finding that inter-party
heterogeneity was a huge factor in how Republicans converted. Curiously, party
heterogeneity in the Baily–Brady sense brought the parties closer in the final vote.
In our case, the role of Democrats is reversed—their agenda is, if not protectionist,
an activist one. Hence, we expect to find exactly the reverse of what Irwin and
4 In a previous version of the chapter, other sectors such as textile and apparel (APPAREL), autos,
aerospace, and other transport equipment (TRANSPORT), resource-based manufactures, for example,
refined petroleum, wood, and furniture (RESOURCE), and other manufactures including machinery,
electronic goods, scientific instruments, primary metals, and iron and steel (OTHERMNF) were used
but they were significant neither statistically nor in magnitude. They were dropped in the analysis
reported here.5 Naturally, all constituency interest variables are candidates for interactions with party. We chose
TARIFF, NTB, and NETEXP as candidates since these are most closely associated with trade
orientation of constituents.
Interest and Ideology in the 1988 Omnibus Trade Act 123
Kroszner find, if their theory about which Republicans convert first is to be
confirmed in our setting—Republicans favoring export interests would be the last
to convert to any protectionist trade legislation that was essentially a Democratic
agenda. A negative sign on the interaction term NETEXP £ PARTY on
protectionist issues and a positive sign on free-trade or export-promoting issues
would be inline with the Irwin–Kroszner mechanism in our setting.
A large literature has developed around the question of whether party measures
interest or ideology in empirical studies. This follows the general finding of the
dominating influence of party in studies of legislative behavior. A consensus, by
no means unanimous, is emerging in the direction of parties as evolving
institutions based on interest. The theory put forth by Cox and McCubbins (1993)
has gained wide acceptance. Cox and McCubbins model party as a mechanism
designed to solve the collective dilemmas of rational but unorganized reelection-
seeking legislators. These collective dilemmas lead to situations that are worse for
all legislators than outcomes attainable by organized action—individual
optimization by self-interested politicians achieve a lower level of success than
if they optimized and their party had a majority. The primary method of solving
collective dilemmas is the creation of leadership posts through which the
organization operates effectively. Hence “parties are invented, structured and
restructured in order to solve a variety of collective dilemmas that legislators
face”. Party, in this view, is an interest-based mechanism designed to resolve the
collective action problem of self-interested actors.6 Parties are groups of
politicians who sometimes act together to achieve collective goals but whose
cooperation is limited by their conflicting preferences. A party that sustains a large
majority in Congress over a long period is able to fashion a portfolio of programs
that favors its constituency. Since activist trade legislation in the House during
the 1970s and 1980s was expected to result in redistributive policies, we feel the
Cox–McCubbins view of parties as interest-based coalitions provide the
appropriate explanation of why the Democratic party was able to form a forceful
coalition during the 1988 Omnibus Act.
The Cox–McCubbins view has been reinforced in recent empirical studies.
Levitt and Snyder (1995) show that parties play a significant role in determining
geographic distribution of domestic federal expenditures. Schickler and Rich
(1997) show that even though there may be dissent within a party, when pitted
6 An alternate view, which was generally accepted before the Cox–McCubbins view became popular,
is the “strong party” model in which parties are effectively treated as unitary actors that maximize some
mix of policy and reelection goals. But that theory was unable to explain how or why a party that
sustains a large majority in Congress over a long period is able to fashion a portfolio of programs that
favors its constituency.
A.K. Bohara and K. Gawande124
against another party, members of the former vote cohesively. McGillivray
(1997) affirmatively tests the hypothesis that the level of party discipline is an
important factor is determining the characteristics of regions which receive
protection. We expect PARTY to enter with a negative sign on votes for
protectionism and activism, certainly on the various versions of the full Bill. We
also expect the interaction of party with constituency interest to be important in
differentiating the Democratic and Republican response.
A counterpoint to the view of party as an interest-based mechanism, is one of
voting based on pure ideology espoused by Poole and Rosenthal (1997). Their
factor-analytic study of historical roll call voting over a hundred years is
interpreted by the authors to lead to the singular conclusion that general liberal-
conservative ideology was the driving force behind voting over the century.
Intuitively, they find one, or at most two, significant factors underlying historical
roll call voting. These factors map into a pattern that split clearly along party lines
and regions, leading Poole and Rosenthal to conclude that the party-region split
represents clear ideological divisions that form the basis of yea and nay voting in
recent history.
Ideology is measured here, as in many studies of voting, by raw ADA scores
(ADA) for House and Senate members.7 We perform a sensitivity analysis of the
effect of including three other measures of ideology (summarized at the end of
Section 6.6). Poole and Rosenthal’s mapping of factors into party and region space
make a compelling case to consider region, after adequately controlling for
interest, as measures of ideology. We indicate three regions with the eponymous
variables NORTHEAST, WEST and SOUTH.
The Poole–Rosenthal interpretation has been challenged by Heckman and
Snyder (1997), who use a simpler factor-analytic method (which the authors show
is more appropriate than the Poole–Rosenthal methodology) to show that six or
more factors are required to fully explain historical voting over the period
considered by Poole and Rosenthal. They put forth the hypothesis that issue-
specific ideology, as opposed to general liberal-conservative ideology, is
responsible for the greater number of dimensions. Issue-specific ideology is not
observable, making it difficult to explicitly assess the Heckman–Snyder claim.
The MVP method provides a natural way of investigating the presence of any
latent factor during legislation of the Omnibus bill via estimates of the error
covariance matrix. Evidence of an issue-specific ideology, here “trade ideology”,
would be reflected in significant error covariances.
7 ADA scores are average ratings by Americans for Democratic Action (ADA) of Congressmen for
1985–1986. The ADA rating is the percentage of times a congressman voted for the ADA position on a
selected sample of 20 issues.
Interest and Ideology in the 1988 Omnibus Trade Act 125
Nollen andQuinn (1994) have performed the only other analysis of voting on this
bill. Their study covers voting onmore than 30 roll calls taken during the legislation
of the Omnibus bill is designed to uncover the coalitions and motivations behind
voting on free trade issues, fair trade issues, strategic trade issues, and
protectionism. Of importance is their use of direct measures of special interest
based on the pressure group model of Olson (1965), which we also employ. They
are the proportion of all PAC contributions accounted by labor PACs
(PACLABOR), domestic corporate PACs (PACCORP), and domestic PACs with
the largest exports or overseas sales (PACINTL).
In sum, the explanatory variables in the voting equations, described fully in
Table 6.3, may be grouped as follows: regional dummies (NORTHEAST,
SOUTH, WEST); committee membership (HTRADESUB, WAYSMEANS,
INTLDEVSUB, SENTRADESUB, FOREIGNREL); special interest pressure
(PACCORP, PACINTL, PACLABOR); constituency interest (UNEMP, FOO-
DEMP, NETEXP); economic interest (TARIFF, NTB); party membership
(PARTY); and inter-party heterogeneity (NETEXP £ PARTY, TARIFF £
PARTY, NTB £ PARTY); and ideology (ADA).
6.5. DATA
Data on roll call votes taken during legislation of the Omnibus bill are from
Congressional Quarterly’s online database. Votes of 99 senators and 425
representatives are used in the study.8 Data on political variables and PAC
contributions are taken fromNollen and Quinn (1994), to which we refer the reader
for data sources and construction. We are grateful to the authors for sharing their
data. Table 6.3 describes the variables, and Tables 6.4 and 6.5 present descriptive
statistics.
Regional dummies for three regions, NORTHEAST, SOUTH, WEST, cover
75% of the Senate and House samples. Committee membership are indicated in
three relevant House committees (HTRADESUB, WAYSMEANS, INTLDEV-
SUB) comprising 10% of reps., and two relevant Senate committees (SENTRA-
DESUB, FOREIGNREL) comprising 36% of senators.
8 Votes other than “Y” (Yea) and “N” (Nay) were treated as follows. “?” (did not vote) were imputed as
1 (Y) or 0 (N), respectively, if the mean on the vote was greater than 0.5 or less, conditioning on party;
“ þ ” were converted to 1 (Y) id “ 2 ” to 0 (N); “#” were converted to 1 (Y) and “X” to 0 (N); “S”
(Speaker did not vote) was deleted; “I” (not allowed to vote for part of 87–88) were deleted. The MVP
method requires the same number of observations for each roll call vote in the model. We included as
many observations as possible via simple imputations. The number of observations for which
imputations were made is generally less than 2% of the sample for any roll call.
A.K. Bohara and K. Gawande126
Table 6.3: Variable definitions.
Variable Description
NORTHEAST 1 if congressman from Northeast, 0
otherwise
SOUTH 1 if congressman from South, 0
otherwise
WEST 1 if congressman from West, 0
otherwise
HTRADESUB 1 if Repp. on Intl. Economic
Policy and Trade Subcommittee of the
House Foreign Affairs Committee
WAYSMEANS 1 if Repp. on Trade subcommittee
of the House Ways and Means
committee 0 otherwise
INTLDEVSUB 1 if Repp. on International Development,
Finance, Trade, and Monetary Policy subcommittee
of the House Banking, Finance, and
Urban Affairs Committee, 0 otherwise
STRADESUB 1 if Senator on Intl. Trade
subcommittee of the Senate Finance committee,
0 otherwise
FOREIGNREL 1 if Senator on Foreign Relations
Committee, 0 otherwise
PARTY 1 if Republican, 0 if Democrat
ADA Liberalism measure rating based on voting
record by Americans for Democratic Action:
0 (very conservative) to 100 (very
liberal)
PACCORP Contributions from domestic corporate PACs as
% of total PAC contributions (domestic
corporations do not include
international corporations defined
for PACINTL)
PACINTL Contributions from international corporate PACs as
% of total PAC contributions
(international corporations
comprise either the 100 largest US
corporations in terms of foreign sales
or the 50 largest US exporters
during any year during 1982–1988
PACLABOR Contributions from labor union PACs as
% of total PAC contributions
UNEMP Unemployed persons as % of the labor force in 1986
FOODEMP Percent of manufacturing labor force in
food processing manufacturing
(continued)
Interest and Ideology in the 1988 Omnibus Trade Act 127
Table 6.3: (continued)
Variable Description
TARIFF Tariff rate as of 1983 (data
section for construction)
NTB Nontariff barrier coverage ratio as of
1983 (see data section for construction)
NETEXT NetExports ¼ Exports 2 Imports, as of 1986 (see data
section for construction)
Notes: data are across states for senators and congressional districts for Representatives. Data (except
FOODEMP, TARIFF, NTB, NETEXP) taken from Nollen and Quinn (1994). See their Appendix A for
more details.
Table 6.4: Descriptive statistics: explanatory variables.
Variable Senate data (N 5 99) House data (N 5 425)
Mean SD Min Max Mean SD Min Max
NORTHEAST 0.222 0.418 0 1 0.247 0.432 0 1
SOUTH 0.263 0.442 0 1 0.292 0.455 0 1
WEST 0.263 0.442 0 1 0.198 0.399 0 1
HTRADESUB – – – – 0.031 0.172 0 1
WAYSMEANS – – – – 0.033 0.179 0 1
INTLDEVSUB – – – – 0.035 0.185 0 1
STRADESUB 0.172 0.379 0 1 – – – –
FOREIGNREL 0.192 0.396 0 1 – – – –
PARTY 0.455 0.500 0 1 0.412 0.492 0 1
ADA 0.467 0.341 0 1 0.478 0.349 0 1
PACCORP 0.294 0.133 0 0.640 0.233 0.115 0 0.672
PACINTL 0.295 0.175 0 0.667 0.259 0.178 0 1.068
PACLABOR 0.144 0.138 0 0.454 0.223 0.205 0 0.838
UNEMP 0.697 0.223 0.280 1.310 0.678 0.263 0.100 1.800
FOODEMP 0.127 0.110 0 0.567 0.129 0.157 0 1
TARIFF 0.538 0.131 0.323 0.916 0.561 0.235 0 1.951
NTB 0.214 0.067 0.123 0.458 0.188 0.110 0 0.596
NETEXP 20.644 0.268 21.538 20.082 20.539 0.444 23.61 0.408
PARTY £ TARIFF 0.232 20.267 0 0.878 0.216 0.290 0 1.952
PARTY £ NTB – – – – 0.072 0.108 0 0.576
PARTY £ NETEXP 20.293 0.376 21.538 0 20.201 0.384 23.61 0.409
Notes: variables are scaled to have similar size for purposes of estimation: PACINTL is scaled by 10,
PACLAB and PACCORP by 100, UNEMP by 10, ADA by 100, TARIFF by 10, NTB by 10. The
descriptive statistics here and the results in subsequent tables reflect this scaling. Mean: sample mean,
SD: sample standard deviation, Min: sample minimum, Max: sample maximum.
A.K. Bohara and K. Gawande128
The ideology versus interest debate is at the heart of the chapter. Ideology is
measured by raw ADA scores for House and Senate members. A sensitivity analysis
also includes three competing measures of ideology: a measure from Levitt (1996)
that purges ADA scores of interest, and two separate ratings based on voting on
economic matters constructed, respectively, by the US Chamber of Commerce and
National Journal (Cohen and Schneider, 1989). Nollen and Quinn’s (1994)measures
of special interest include the proportion of all PAC contributions accounted by labor
PACs (PACLABOR), domestic corporate PACs (PACCORP), and domestic
corporate PACs with the largest exports or overseas sales (PACINTL).9
The unemployment rate (UNEMP) is from Nollen and Quinn. The share of the
Food processing sector in manufacturing employment (FOODEMP) is con-
structed as follows. The state level data on employment were straightforwardly
obtained from the geographic area series of the 1987 Census of manufacturing.
The mapping in Congressional District Atlas (Department of Commerce, 1984)
was used to concord the county-level Census data into districts.10
Data on ad valorem tariffs (TARIFF), nontariff barrier coverage ratios (NTB),
and net-exports-to-shipments ratio (NETEXP) are unavailable across states and
districts, and are estimated using Conybeare’s method (also used by Bailey and
Brady (1998)) from their national averages. First, US averages for exports ðXÞ;
Table 6.5: Descriptive statistics: roll calls.
Senate data House data
Roll call (Year, #) Mean SD Roll call (year, #) Mean SD
1987, 158 0.73 0.45 1987, 72 0.50 0.50
1987, 175 0.56 0.50 1987, 77 0.37 0.48
1987, 178 0.42 0.50 1987, 78 0.67 0.47
1987, 179 0.72 0.44 1987, 426 0.41 0.49
1987, 208 0.73 0.44 1988, 66 0.74 0.44
1988, 110 0.64 0.48 1988, 231 0.89 0.31
1988, 285 0.73 0.45
SD ¼ Sample standard deviation.
9 Nollen and Quinn (1994, Appendix A) report that data for senators are from 1983–1988 and for reps.
from 1985–1986. In 1987–1988 there were 1873 corporate PACs and 135 International PACs in
1987–1988. International PACs are not counted as domestic corporate PACs in computing PACCORP.
The total of the three PAC variables sums, on average, to around 60% of the total PAC money received
by senators and reps.10 If the county-to-district mapping was one-to-many, the employment data were proportioned equally;
if the mapping was many-to-one, then the county data were added up. The employment shares variables
were then computed.
Interest and Ideology in the 1988 Omnibus Trade Act 129
imports ðMÞ; shipments ðSÞ and value added (V) all for 1986, and TARIFF and
NTB as of 1983,11 are computed for the 20 two-digit SIC industries. The two-digit
values are then weighted down to a state/district level value, using as weights the
proportions of the state/district’s manufacturing value added accounted by the 20
industries. For example, the tariff rate on imports into state or district i is given as
TARIFFi ¼P20
j¼1 wijTARIFFj; where wij ¼ Vij=Vi:The variables NTBi; ðX=SÞi; ðM=SÞi; are similarly constructed as weighted
averages of their two-digit values using wij as weights. The net export to shipment
ratio (NETEXP) for state/district i is then computed as ðX=SÞi 2 ðM=SÞi:
6.6. EMPIRICAL ANALYSIS
6.6.1. Substantive issues of focus
Linear dependencies among roll call votes may make it difficult to estimate the
MVP model. For this reason, MVP models of House and Senate voting include at
most three roll calls at a time. The choice of which roll calls to include is based on
two issues common toHouse and Senate voting and a third issue exclusive to Senate
voting.12 The first and most important set of issues concerned voting on the full
Omnibus bill itself. There were four House bills at different stages of the process
(#78—House passage, #66—adoption of conference bill, #150—veto override,
#231—final passage), and four Senate bills (#208—Senate passage, #110—
adoption of conference bill, #16—veto override, #288—final passage).13
The second set of issues was related to a push for protectionism. Action on the
House floor centered around the controversial Gephardt amendment (#72), which
explicitly targeted Japan, Taiwan, and Hong Kong. It barely passed the House,
overcame two efforts at getting rid of it (#77, #426), but was dropped in conference.
11 X;M; S; V are from the 1986 Annual Survey of Manufactures. TARIFF, NTB are described in Trefler
(1993).12 The bills are chosen to represent different clusterings of issues. Otherwise the multivariate analysis
of very similar bills breaks down due to linear dependency among them. Narratives from weekly issues
of Congressional Quarterly (CQ), examination of the nature of the bill (protectionist, free, curtailment of
Presidential powers, strategic), voting splits, breakdown of the vote by party, and simple correlations are
inputs into deciding how the bills cluster.13 There were no explicit votes taken on many important characteristics that were bundled into the
Omnibus bill. The Super 301 provisions under Section 301, which granted Congress the power to
undertake unilateral action against unfair trade practices by any partner, was one such key
characteristic. It was considered by free-traders as the single most damaging aspect of the Act (see, e.g.,
Bhagwati and Patrick, 1990). Since no roll call votes were taken on 301, the only analysis on this issue
is from the voting on the passage of successive versions of the bill itself.
A.K. Bohara and K. Gawande130
On the Senate floor a set of bills relating to protectionism sought to amend Section
201 dealing with import injury cases (#158, #179, #206), with some winning and
others losing. The third set of issues, debated in the Senate, concerned discretionary
powers of the executive which they had used to overturn decisions by the ITC
(#175, #178, #285). Championed by Bentsen (D-TX), Chair of the Senate Finance
Committee, those issues also concerned the transfer of authority over protection
cases from the President to the US Trade Representative (USTR). The USTR was
perceived as being more independent of state agencies under control of the
executive, who could interfere with trade cases.
6.6.2. Analysis of MVP estimates from HOUSE voting
Estimates from MVP models analyzing House voting is presented in Table 6.6.14
Model 1 estimates the trivariate model of voting on Gephardt’s (D) amendment to
the House Omnibus bill (#72), Michel’s (R) somewhat less protectionist substitute
(#77) for the House Omnibus bill, and passage of the House Omnibus bill (#78).
Model 2 jointly analyzes the Michel (R) motion to drop Gephardt (#426) from the
conference bill, and voting on the conference bill (#66) itself. Model 3 jointly
analyzes passage of the House version (#78), Conference version (#66), and the
final version (#231) signed into law. Each of the three MVP models includes a
version of the full Omnibus bill for the reason that the important roll call votes were
amendments to the Omnibus bill under consideration, and helped shape the bill.
The coefficient estimates fromModel 1 testifies to the persistence of the ideology
versus interest debate. The evidence across voting on the Gephardt (D) amendment
(#72), the Michel (R) substitute (#77), and the House version of the Omnibus bill
(#78) show that while the ADA and regions (if one subscribes to the Poole–
Rosenthal view of region as representing ideological tradition) are significant deter-
minant of voting on protectionism, special interests (in particular PACLABOR),
constituency interests (UNEMP, FOODEMPL), and other measures of interest
(TARIFF, NTB) are each statistically significant across the three roll calls.15
14 We are indebted to Sid Chib for providing us the code for the MVPmodel estimation. After a bum-in
of 500 discarded samples, a sample of 5000 was used to compute the estimates. The estimates are
comparable to single-equation probit results, but there are differences in statistical significance and
magnitude due to systems estimation and the inclusion of S in the MVPmodel. Computational note: on
a P4 machine, a trivariate House model such as Model 1 in Table 6.6 took 45 min and a trivariate
Senate model took 15 min to run.15 Asterisked estimates are interpreted as being statistically significant. * implies that 2 . ltl $ 1 and
** implies ltl . 2:A rationale for using this criterion to denote statistical significance is that (in a linear
regression setting) only if a variable has a coefficient with ltl $ 1 will its exclusion lower the model’s
adjusted R2 and its inclusion raise it (e.g., Greene, 1993, p. 193).
Interest and Ideology in the 1988 Omnibus Trade Act 131
Table 6.6: Estimates from MVP models of House voting.
Variable Model 1 Protectionism 1 RepublicanBill 1 House Bill
Model 2 GephardtRedux 1 Conference Bill
Model 3 House 1 Conference 1 FinalBill
# 72 (P) # 77 (F) # 78 (P) # 426 (F) # 66 (P) # 78 (P) # 66 (A/P) # 231 (A)
CONSTANT 2 1.820* 2.874* 2 1.284* 1.835** 0.058 2 1.323* 0.585 20.114
NORTHEAST 0.315* 2 1.852** 0.113 20.389* 0.506* 0.237 0.670* 0.444*
SOUTH 20.669** 20.074 20.303 0.265 0.286* 20.252 0.373* 1.207**WEST 2 1.657** 0.705* 2 2.325** 0.925** 20.302* 2 1.797** 20.325 20.105
HTRADESUB 0.115 20.091 1.094* 20.326 1.489** 0.764* 1.713** 0.599
WAYSMEANS 20.861* 0.989* 0.722* 20.308 2 2.476** 0.530* 21.936** 20.124
INTLDEVSUB 0.672* 1.047 0.157 0.737* 20.617* 0.029 20.685* 0.466
PARTY 2 1.543** 4.013** 2 1.453* 2.078** 20.089 2 1.433* 20.527 0.043
ADA 0.577* 2 3.119** 2.769** 2 1.841** 3.357** 2.485** 3.116** 1.866**PACCORP 1.982* 2 6.393** 0.779 21.929* 22.570* 1.440* 22.162* 22.056*
PACINTL 20.304 0.422 20.229 20.255 20.023 20.428 20.335** 0.363
PACLABOR 3.190** 2 2.358* 0.721 2 3.124** 2 1.945* 0.888 2 2.293 2 1.445*UNEMP 1.256** 20.941* 0.869* 21.198** 0.722* 0.678* 0.443* 1.132**FOODEMPL 22.225** 6.776** 2 3.820** 2.833** 0.899 2 2.998** 1.115* 0.793
A.K.Bohara
andK.Gawande
132
TARIFF 0.692* 2 3.840* 2.268* 0.064 0.704 2.703* 0.957 1.856*NTB 2.964** 2 7.342** 2 7.146** 2 5.261** 6.868* 4.654** 4.900* 20.934
NETEXP 0.070 2 1.209 0.717 20.550* 20.646 0.811* 20.149 2 0.654PARTY £ TARIFF 0.022 1.441 20.075 21.274* 2 1.617 20.437 2 1.839 2 1.221PARTY £ NTB 20.278 0.321 2 4.251* 2.096 2 7.777* 23.279** 2 5.833* 20.884
PARTY £ NETEXP 0.210 20.044 20.308 20.035 0.518 20.497 20.038 0.451
{s12;s13;s23} {20.907**, 0.833**, 20.836**} {20.578**, –, –} {0.588**, 0.612**, 0.801**}
N 425 425 425
K 19 19 19 19 19 19 19 19
Maddala’s R2 0.751 0.680 0.715
Notes: posterior estimates based on weak priors using Chib and Greenberg’s (1998) method. Results based on 5000 samples after discarding 500
initial samples. All prior locations set to zero and prior standard errors set to 10. **(posterior) ltl . 2 and *2 $ ltl . 1: Description of roll calls in
Tables 6.1 and 6.2. Bold indicates that beta coefficient is greater than 1. The beta coefficient is equal to coefficient £ SDðxÞ=SDðyÞ; where SDðxÞ is thesample standard deviation of the independent variable and SDðyÞ that of the dependent variable. Beta coefficient . 1 implies that a one standard
deviation change in x induces a greater than one standard deviation change in y: Frequentist measures of fit: Maddala’s R2 ¼ 12 ðL0=LFÞ2=N; whereL0 and LF are, respectively, the data likelihoods of the null model (just intercept terms) and the full model, and N is the number of observations. On a
trade-activism scale of 0–10, with 0 denoting free trade and 10 denoting extreme protectionism, (F) denotes a score in the interval ½0; 5Þ; and (P)
denotes a score in the interval ½5; 10�: These subjective scores fit some but not all bills. Hence, (A) denotes an activist bill that allows strategic
unilateralism and perhaps protectionism. {s12;s13;s23} are the upper triangular elements of the correlation matrix in the trivariate model.
InterestandIdeologyin
the1988OmnibusTradeAct
133
Coefficients in bold are significant in magnitude in the sense that their standardized
beta coefficients exceed unity, that is, a one standard deviation change in the
explanatory variable induces a greater than one standard deviation change in the
dependent variable in the direction of the sign on the coefficient.16 Of all the
variables, PARTY dominates in beta coefficients (21.52, 4.09,21.52, across the
three votes). Whether ideology dominates interest depends if one subscribes to the
Cox–McCubbins view of PARTY as representing interest or the Poole–Rosenthal
view of PARTY as representing ideology. Since we explicitly include ADA scores
as measures of ideology, we come out in favor of the Cox–McCubbins view. Of
course, a formal model comparison, which we report presently, is required before a
definitive statement about interest versus ideology may be made.
Since the econometric specification is motivated by the models set out in the
theory section, the estimates are well suited for a test of those models in the
present setting. Conybeare’s hypothesis that lower opportunity costs of providing
protection will lead to pro-protectionist voting is verified by the statistically
significant coefficients on UNEMP with the expected signs (positive for
protectionist votes #72 and #78 and negative for the less protectionist vote
#77). However, none of the coefficients on UNEMP is strikingly large in
magnitude. Measures of existing protection (TARIFF, NTB) have the expected
signs and are statistically precise as well as large in magnitude, thereby strongly
affirming Conybeare’s demand side hypothesis that protection begets protection.
Quite unsurprisingly, his hypothesis about PARTY is also strongly affirmed. What
is noteworthy is that Conybeare’s theory of why party is important is different
from, say, Cox and McCubbins. Party in Conybeare’s model is a technology
parameter in a politician’s voting production function. Just as technology defines
the “capacity” to find cost-minimizing combination of factors to supply a good, so
too party defines the institutional capacity of politicians to supply protection.
Since the legislative push for the trade bill stemmed from the Democratic majority
in both chambers, Democrats possessed a larger institutional capacity to vote for
trade activism. Essentially, it is a theory about neoclassical optimization, and
hence one based on interest.
There is strong evidence in favor of the Olson pressure group hypothesis from
the large and precisely measured estimates on PACLABOR and PACCORP.
There is weak evidence on the Baily–Brady hypothesis of inter-party
heterogeneity from the statistically significant coefficient on PARTY £ NTB on
vote #78 (House passage). Significance of the interaction term implies that the
16 Statistical significance implies precision in measuring the coefficient, while large beta coefficients
imply significant in magnitude. Neither implies the other, yet, both are worth reporting because they
each connote “importance” according to a variety of criteria.
A.K. Bohara and K. Gawande134
political response to existing NTB protection was heterogeneous across the two
parties. For the #78 vote on House passage, ›(vote)/›NTB ¼ 7.146 2 4.251 £
PARTY. For Republicans (PARTY ¼ 1), ›(vote)/›NTB ¼ 2.895, and for
Democrats (PARTY ¼ 0) it is 7.146. Though existing NTB protection led to
yea voting on passage regardless of party, the probability of a “yea” vote due to
any increase in NTB (across regions) was 7:146=2:895 ¼ 247% higher for
Democrats than Republicans.
A feature of Model 1, indeed across the three models, is the statistically
significant and large estimates on elements of S; the error correlation matrix.
Methodologically, there are efficiency gains in MVP estimation compared with
single equation methods. This is demonstrated in standard econometrics texts (e.g.,
Greene, 1993) in a linear, seemingly unrelated regression (SUR) setting, and a
similar analysis applies to the probit case. Indeed, we should expect voting on
similar issues within a short period of time to display such correlations. Yet, since
each separate vote contains some new information, inference should take into
account correlations across votes. Substantively, the error correlations imply the
existence of a common underlying latent factor responsible for the high correlations
among the votes. While we are inclined to think of this as Heckman and Snyder’s
(1997) “issue-specific ideology”, the high-error correlations among related votes
may also be due to intra-party logrolling coalitions as well as inter-party vote
trading (see, e.g., Stratmann, 1992). Since logrolling is intrinsic to Omnibus bills
generally, a full investigation of logrolling versus issue-specific ideology is
required before a definitive answer may emerge. That would take us beyond the
scope of this chapter, and we leave this interesting issue open to future research.17
Gephardt had decided to take his amendment directly to the House-Senate
conference. Model 2 analyzes voting on a motion by Michel (R) (#426) to instruct
House conferees to eliminate it in conference, jointly with voting on the
conference version of the Omnibus bill (#66). Interestingly, even though the
Democrats knew that Gephardt’s amendment was effectively dead, they defeated
Michel’s motion 175–239 to show party unity in the face of a Republican attack.
The Omnibus bill adopted in conference 312–107 (Gephardt was one of two
17 Extracting this trade-specific ideology through factor analysis of the errors may be done, say, for
Model 1, as follows. Extract the n £ 3 matrix of residuals given the estimated coefficients and the
covariance matrix S: Perform a factor analysis or principal component analysis on the matrix, test to
see how many significant factors exist. If one factor is adequate, “test” to see whether this is
interpretable as trade ideology by regressing on a set of purely trade related votes (other than the votes
explained in Model 1) and examining the coefficients and the regression’s explanatory power. A full-
blown LISREL structural analysis, which encompasses the factor analytic method as a special case, is
worth undertaking in order to understand the trade ideology component. Bohara et al. (2003) analyze
the raw voting data using such a method.
Interest and Ideology in the 1988 Omnibus Trade Act 135
Democrats voting nay) did not contain Gephardt, and appeared less protectionist
than the House-passed version in other respects as well. Expectedly, the pattern on
the coefficient estimates for voting on Michel’s motion is very similar to earlier
voting on Gephardt’s amendment (#72), but with the signs reversed.
The coefficient estimates for voting on the conference version of the Omnibus
bill (#66), on the other hand, are notably different from the estimates for voting on
the House passage of the bill (#78). Might this mark the beginning of a Irwin–
Kroszner conversion of Republicans? The estimates of the interaction PARTY £
NETEXP has the right sign but is imprecisely measured, and is not significant in
magnitude (beta coefficient ¼ 0.454). The conversion that did occurwas bySouthern
representatives in favor of this less protectionist version, and a melting of the strong
opposition byWestern representatives to theOmnibus bill. If one considers region as
ameasure of ideology, then ideology did play a role in the conversion in theHouse.18
Model 3 jointly analyzes voting on the House bill (#78), Conference bill (#66),
and the version signed into law after overcoming the veto (#231). The motivation
for estimating thismodel is to focus on characteristics of the bills that were common
across the three versions but for which no motions were made and no roll calls
taken. They include major Section 301 (unfair trade) amendments which led to the
Super 301 provision permitting unilateral actions by the US against unfair foreign
practices, transfer of authority over Section 301 cases to a lower level decision
making body—the USTR—and extensions of antidumping/countervailing duty
(AD/CVD) cases to high-tech and agricultural goods. The bundling of these
amendments made for an aggressive bill, and gained the support of a sufficient
number of Republicans at each stage, passing each version overwhelmingly (290–
137, 312–107, 376–45, respectively).
Important determinants of voting for the full bills were (i) liberal-conservative
ideology (ADA), (ii) labor PAC contributions, (iii) existing tariff and NTB
protection, and (iv) inter-party heterogeneity measured by interacting party with
tariff and NTB. Southern and Western Republicans converted with the dropping
of the labor-backed plant closing provision to overturn the veto (#231). Hence,
the importance of PARTY diminished towards the end, as the bill became more
bi-partisan.
18 A key difference between #78 and #66, other than the absence of Gephardt-like provisions in
#66, was that the conference bill contained a labor-backed plant closing amendment that made it
mandatory for employers to give a 60-day notice prior to any plant closings or layoffs. It is,
therefore, surprising that PACLABOR has a negative coefficient, indicating that representatives
supported strongly by labor PAC money voted against it. Since liberal representatives with high
ADA scores are also those strongly supported by labor PAC contributions we suspect ADA is
picking up the effects of labor PAC contributions. Due to the high correlation between ADA and
PACLABOR, the weaker PACLABOR variable shows up with a contrary sign.
A.K. Bohara and K. Gawande136
6.6.3. Analysis of MVP estimates from SENATE voting
In the analysis of Senate voting our focus is on three issues central to the Senate
debate: (i) transfer of authority over trade cases from the President to the USTR, (ii)
reducing Presidential discretion to overrule decisions made by the ITC, and
(iii) Section 201/301 amendments regarding injury cases, and unfair trade practices.
Bentsen had emphasized transfer of authority and presidential discretion as the
key features of the Senate bill. The first roll call came early in the debate in the
form of Packwood’s (R-OR) amendment (#178), embodying those twin questions.
The amendment would have allowed the President to exercise discretion in ruling
on those cases where the ITC requested the President to consider national
economic interests. The Senate turned back the amendment 41–55 in a party line
vote (with at least two-thirds from each party voting on either side). While
debating the final version of the Omnibus bill, Bentsen easily fought off an
amendment (#285) that would restore authority to the President, 69–26, with 40%
of participating Republicans voting with Bentsen.
In Table 6.7, Model 1 presents MVP estimates from the trivariate model of
voting on these two “transfer of authority” votes and the conference version of
the bill (#110), which came between the two roll calls.19 ADA scores become
important only on voting on the bill (#110) but not on the issue of transfer of
authority (#178, #285). However, region, in particular SOUTH and WEST are
precisely measured and important in magnitude to Senate voting on this issue.
Whether this is indicative of (traditional) liberal-conservative ideology or trade
ideology due to the generally pro-free-trade stance of Southern and Western
senators is key to the question of whether ideology is at all important to
Senate voting on trade. Interestingly, it was easier to get the Southern senators to
vote for the conference bill (#110) than Western senators, who voted against it.
Conybeare’s model holds with less force in Senate voting on the transfer of
authority issue than in House voting. Existing protection (NTB, TARIFF) force-
fully leads to nay voting on Packwood’s amendment (#178), but is not significant
for voting on the conference version of the Omnibus bill (#110). Though not
statistically significant, NETEXP is significant in magnitude. The coefficient on
PARTY £ TARIFF indicates that evidence on the party heterogeneity hypothesis
of Baily and Brady is significant in magnitude. In the case of voting on the
conference bill (#110) it is statistically significant as well. The Olson pressure
group hypothesis is probably the most clearly vindicated model. The PAC
19 Eighteen variables is quite a burden for the 99 Senate observations to bear in estimation of the MVP
model. The PARTY £ NTB variable was dropped to reduce the number of regressors and facilitate
estimation.
Interest and Ideology in the 1988 Omnibus Trade Act 137
Table 6.7: Estimates from MVP models of Senate voting.
Variable Model 1 Transfer of authority 1
Conference BillModel 2 Presidential Discretion 1
Senate BillModel 3 Section 201/301 1 Senate
Bill
#178 (F) #285 (P) #110 (A) #175 (F) #178 (F) #208 (A/P) #158 (F) #179 (P) #208 (A/P)
CONSTANT 7.280** 2.519 2 3.250 2 1.342 7.015** 1.331 7.582** 2 2.949* 1.275NORTHEAST 20.167 1.343* 2.199* 2.392** 0.188 20.027 20.257 20.263 20.740
SOUTH 3.169** 2 3.707** 2.812** 2 1.630** 2.591** 1.651* 3.391** 20.434 1.054WEST 2.318** 2 2.55** 2 2.789* 20.437* 2.128** 2 2.892** 0.385 20.637* 2 2.978**STRADESUB 2 0.730* 2.474** 4.857** 0.103 20.446 3.188* 0.683* 20.24 2.881**FOREIGNREL 1.516** 0.844* 2 1.412* 0.829* 1.417** 2 1.437 20.314 20.159 2 1.351*PARTY 6.694** 2 7.023* 2 2.884 1.029 4.916** 2 7.111* 2.564* 2 3.741* 2 6.638*ADA 0.796 20.890 4.247* 0.264 0.103 5.582** 0.697 2 2.490** 4.948**PACCORP 1.961 4.959* 11.39** 21.455 0.927 3.141 21.814 22.52 2.257
PACINTL 2 9.339** 2 7.552** 3.818* 22.195* 2 7.526** 3.717* 22.028* 3.979** 3.465*PACLABOR 2.023 2 1.006 9.847** 0.031 1.739 0.112 22.714* 3.088* 2.205
UNEMP 2 2.950* 4.118* 21.050 1.815* 21.896* 2 3.740* 2 3.217** 4.853** 2 3.311*
A.K.Bohara
andK.Gawande
138
FOOD 2.791 7.729* 3.341 3.734* 3.391 16.25** 6.124** 0.002 13.08**TARIFF 2 11.59** 2.183 22.662 20.543 2 11.01** 4.251 2 7.541 2.393 2.565
NTB 2 7.866* 22.298 20.733 2 10.07** 2 7.824* 22.215 2 7.427** 6.104* 20.782
NETEXP 1.829 2 3.621 2 4.068* 2 3.271** 2.112* 21.232 20.830 1.643* 2 5.512PARTY £ TARIFF 2 1.963 2 1.991 2 9.687* 6.423** 21.328 21.196 2 2.826 1.269 1.336
PARTY £ NETEXP 2 1.216 21.176 20.475 2.662 2 1.714 20.827 0.928 2 4.433** 3.763{s12;s13;s23} {0.713**, 20.352*. 0.496*} {0.618**, 0.146, 20.358*} {20.624**, 20.465**, 0.031}
N 99 99 99
K 18 18 18 18 18 18 18 18 18
Maddala’s R2 0.826 0.815 0.761
Notes: posterior estimates based on weak priors using Chib and Greenberg’s (1998) method. Results based on 5000 samples after discarding 500
initial samples. All prior locations set to zero and prior standard errors set to 10. **(posterior) ltl . 2 and *2 $ ltl . 1: Description of roll calls in
Tables 6.1 and 6.2. Bold indicates that beta coefficient is greater than 1. The beta coefficient is equal to coefficient £ SDðxÞ=SDðyÞ; where SDðxÞ is thesample standard deviation of the independent variable and SDðyÞ that of the dependent variable. Beta coefficient . 1 implies that a one standard
deviation change in x induces a greater than one standard deviation change in y: Frequentist measures of fit: Maddala’s R2 ¼ 12 ðL0=LFÞ2=N; whereL0 and LF are, respectively, the data likelihoods of the null model (just intercept terms) and the full model, and N is the number of observations. On a
trade-activism scale of 0–10, with 0 denoting free trade and 10 denoting extreme protectionism, (F) denotes a score in the interval ½0; 5Þ; and (P)
denotes a score in the interval ½5; 10�: These subjective scores fit some but not all bills. Hence, (A) denotes an activist bill that allows strategic
unilateralism and perhaps protectionism. {s12;s13;s23} are the upper triangular elements of the correlation matrix in the trivariate model.
InterestandIdeologyin
the1988OmnibusTradeAct
139
contribution variables are significant statistically as well as in magnitude. All three
types of PAC contributions strongly encouraged voting in favor of the conference
bill. Theories that focus on the role of committees in congressional organiza-
tion (e.g., Weingast–Marshall (1988), Shepsle–Weingast (1987), Weingast–
Shepsle–Johnson, Cox–McCubbins) also receive strong support from the size and
signs on the coefficients on STRADESUB and FOREIGNREL. A final noteworthy
feature of this model is that though the influence of party dominates the effect of
most other coefficients in magnitude. In Model 1 PARTY has beta coefficients
{6:75;27:85; and22:99}; larger than any other coefficient on the first two votes,and surpassed only by PARTY £ TARIFF on the third vote.
The issue of Presidential discretion, analyzed in Model 2 was linked with the
issue of transfer of authority. Voting on Packwood’s amendment (#178), which
would effectively transfer discretion to the President, appears in Model 2 as well.
The other roll call vote on the issue of discretion was an amendment by Bradley
(D-NJ) and Packwood to remove from the bill a proposal requiring the President
to keep oil imports below 50% of consumption. The Bradley/ Packwood motion
was interesting because it combined the conflicting issues of increasing
Presidential discretion (opposed by Democrats, supported by Republicans) and
removing protection to a sector with strong linkages to other sectors (supported
by downstream industries, supported by environmental groups, and opposed by
oil interests and senators from oil producing constituencies). The motion won
55–41. Since roll calls #175 and #178 sought to amend the Senate version of the
Omnibus bill (#208), Model 2 jointly estimates the three votes.
In this trivariate model, region, specifically SOUTH and WEST are again
important, though as in the first model, it was easier to get Southern senators to vote
for the Senate bill (#208) while Western senator voted against it. The Bradley/
Packwood amendment (#175) presents an interesting experiment to test how
strongly region in fact represents ideology. If so, then in this instance interest got
the better of ideology. While the amendment was strongly supported by Eastern
senators who represented interests of parties downstream of oil producers, it was
strongly opposed by Southern senators representing oil producers. Party continued
to be important inmagnitude. It ismeasuredwith less precision especiallywhere the
interaction terms are measured with significance. Existing protection (TARIFF,
NTB) and net exports (NETEXP) continued to be important on the issue of
Presidential discretion (#175, #178), though not in voting for the Senate bill (#208).
A comparison of the Senate version (#208) of the bill and the conference bill
(#110) (see Model 1) illuminates the role of interest. The major difference
between the two bills was that the latter contained the plant closing provision that
led eight Republicans who had supported the Senate version to vote against the
conference version. Hence, while #208 did not engender the support of labor
A.K. Bohara and K. Gawande140
(PACLABOR), #110 did. States with a greater proportion of employment in food
processing, who had earlier supported the Senate version #208, dropped their
support on account of the plant closing provision.
Model 3, on the issue of amendments to Section 201 (injury from imports) and
301 (unfair trade practices), probably brings out the importance of constituency
interests best. Many traditional sectors had successfully brought injury cases to
the ITC, and it had been the workhorse for delivering protection through
antidumping/countervailing duty (AD/CVD) measures for the past two decades.
The very first roll call vote taken was on this issue (#158) was Bentsen’s attempt
to kill Specter’s (R-PA) protectionist amendment that would allow private right
of action by injured domestic producer. Bentsen won 69–27 in one of the few
votes with bipartisan support. #179 was a moderately protectionist amendment
by Domenici (R-NM) to expand the definition of unfair trade practices to include
excessive government subsidies by foreign governments. It was voted in by
another bipartisan vote, 71–28. Model 3 estimates these two votes jointly with
voting on the Senate bill, which soon followed the two votes.
Party, ideology, special interests, and constituency interests are each significant
determinants of Section 201/301 votes (#158, #179). The importance of existing
protection, unemployment, and employment in food processing affirmConybeare’s
hypothesis in the context of Senate voting on #158. The significance on the
NETEXP £ PARTY in #179 attest to inter-party heterogeneity in response to net
exports: Republicans from states with high net exports were prone to vote against
protection, even were such as amendment proposed by a fellow Republican.
6.6.4. Sensitivity analyses
The models reported in Tables 6.6 and 6.7 are estimated with substitutions and
additions of variables as follows. First, we estimated the Senate models, replacing
the raw ADA scores with the refined ADA scores in Levitt (1996), where the raw
scores are purged of influences other than ideology. Hence Levitt’s refinement
purports to measure pure ideology.20 The findings are in general similar to what
we have reported. In some instances, however, the refined scores are far less
20 Since Levitt’s data are available for only 53 of 100 senators of the 100th Congress, the sample must
be either truncated to include only the 53 senators, or missing values must be imputed for other senators
based on their raw ADA scores. We chose the latter course. Levitt indicates a correlation coefficient in
excess of 0.90 between raw ADA scores and his estimated ideology measure, which provides a basis for
the imputations. Using the sample of 53 senators we estimate an auxiliary regression of Levitt’s scores
on raw ADA scores. The estimated regression is then used to predict missing Levitt scores for the
remaining 48 senators. The adjusted R2 on the fitting equation is 0.935.
Interest and Ideology in the 1988 Omnibus Trade Act 141
significant, statistically as well as in magnitude, for example, in Model 3 for
Senate voting.
We next estimated each of the House and Senate models, with two different
measures of ideology, one at a time, in place of ADA. These ratings are based
on prior voting on economic issues, one constructed by the US Chamber of
Commerce (called CCUS) and another constructed by Cohen and Schneider
(1989) in the National Journal (called CONECON). Surprisingly, both measures
(which are quite correlated) turn out to be far less important to both, House and
Senate voting, than our estimates with ADA show. This finding is significant for
opponents of ideology as a major explanation of voting, because it may be argued
that CCUS and CONECON are appropriate ideology measures for understanding
voting on redistributive issues, while ADA may be more appropriate for under-
standing social issues (which comprise the majority of historical roll call votes
studied by Poole and Rosenthal). If CCUS and CONECON are in fact the
theoretically appropriate measures of ideology, then our findings do not support
ideology as a strong basis for voting on redistributive issues.
It is instructive at this point to discuss results from the innovative study of the
same Act by Nollen and Quinn (1994). Nollen and Quinn use a logistic regression
to explore the effects of political institutions, ideology, and economic conditions
on the outcomes of all 30 Senate and 20 House trade roll calls of the 100th US
Congress. Specifically, they assess the conditions under which members of the
100th Congress support free trade, fair trade, strategic trade, or classical
protectionism. Further, Nollen and Quinn explore whether different types of trade
intervention (i.e., fair trade and strategic trade) are different in political practice
from classical protectionism. If so, conditions that lead members of Congress to
favor or oppose different types of intervention will differ. To explore their idea,
Nollen and Quinn classify the 30 Senate and 20 House bills as fair trade, strategic
trade or classical protectionism (eight of the 30 Senate votes are classified as fair
trade, three as strategic trade, and 19 as classical protectionism; of the 20 House
votes, seven are classified as fair trade, two as strategic trade, and 11 as classical
protectionism.)
Nollen and Quinn find strong evidence that all types of trade intervention are
opposed by the Republican Party; however, they also find convincing evidence
that fair and strategic trade policies are not protectionism in disguise. For instance,
Senators who receive a relatively high amount of contributions from international
corporate PACs tend to oppose protectionist policies but favor fair trade. Votes
on protectionist roll calls are primarily explained by economic variables such as
geographic region, export employment, unemployment rates, or PAC contri-
butions. Fair or strategic trade polices are primarily explained by political
variables including party membership (recall, Republicans oppose all forms of
A.K. Bohara and K. Gawande142
trade intervention), ADA scores and committee memberships, and with a lesser
extent by economic variables. For example, protectionism opposed by Westerners
(or supported by members from the South or East), or when export employment in
a member’s district or state is relatively high. Protectionism is favored by
members of the House who receive greater amounts of PAC contributions from
domestic or labor union PACs (opposed by those who receive greater amounts of
contributions from international PACs), or are from districts with higher rates of
unemployment. Across both houses, members that tend to be more liberal (as
measured by ADA scores) favor fair and strategic trade policies. Trade committee
membership has varying outcomes on trade policy legislation. Members of the
Senate Foreign Relations Committee oppose all types of intervention, while
members of the Senate Trade Committee oppose classical protectionism.
Members of the House International Trade Subcommittee tend to favor fair and
strategic trade policies, while members of the House Ways and Means Trade
Subcommittee oppose fair and protectionist trade policies.
In comparison, our multivariate results yield the following differences. In
House voting, their single equation logistic estimates are different from ours for
UNEMP (#66, #231, #77), PACINTL (#66, #77), PACCORP (#231, #78),
PACLABOR (#66, #78) and WAYSMEANS (#78). In Senate voting, we differ
on ADA (#208, #179), PACINTL (#178, #175, #179), PACCORP (#158, #178,
#208), PACLABOR (#208, #110 #175, #285). However, our results generally
agree on PARTY and ADA. By and large, our results support the suggestion
from Nollen and Quinn’s findings that trade policy is probably not unidimen-
sional; members’ ideological views on various trade policies (fair trade, strategic
trade, etc.) may be different.
6.7. MODEL COMPARISONS: IDEOLOGY VERSUS INTEREST
The traditional method of inference used in the previous section of examining the
size and sign on individual coefficients, led to the conclusion that although the
determinants of voting varied across House and Senate, there were enough
instances inwhich each of a number of theories received significant support. This as
it should be.We have examined a number of distinct votes covering a broad range of
issues. In general, a hypothesis that a (common sensical) theory has no explanatory
power at all (i.e., the coefficients on variables representing the theory are all strictly
zero) can be rejected without recourse to data. A more interesting, and more
demanding, line of enquiry is how important a theory relative to an alternative one.
This exercise requires the testing of theories in the presence of well-specified
Interest and Ideology in the 1988 Omnibus Trade Act 143
alternatives. In this section we present Bayesian model comparisons of alternative
theories, focusing particularly on the comparison of ideology versus interest.
The classical methodology of non-nested tests is somewhat ambiguous. There
are at least three different asymptotic methods for non-nested model comparisons,
each based on a different philosophy.21 In contrast, the Bayesian method is
conceptually straightforward (see, e.g., Zellner, 1971). The posterior probabilities
of two non-nested models Mi; i ¼ 1; 2 are given by
PðM1lyÞ ¼Pð ylM1ÞPðM1Þ
Pð ylM1Þ þ Pð ylM2Þ; and
PðM2lyÞ ¼Pð ylM2ÞPðM2Þ
Pð ylM2Þ þ Pð ylM1Þ;
ð2Þ
where PðylMiÞ is the marginal likelihood of the data y given modelMi; and PðMiÞ
is the prior (subjective) probability of model i: Their relative likelihood is
PðM1lyÞ
PðM2lyÞ¼
f ðylM1Þ
f ðylM2Þ
PðM1Þ
PðM2Þ: ð3Þ
Chib’s (1995) method makes it possible to estimate the marginal likelihoods from
the output of the MCMC simulations. In all the model comparisons, the prior odds
PðM1Þ=PðM2Þ in Equation 3 are set to unity to represent nonsubjectivity. Readers
may update the results to take into account their prior odds.
Our aim is to discriminate among hypotheses of ideology versus interest, while
giving each hypothesis the opportunity to put their best foot forward. Define
the following full models for the House and Senate, respectively, as comprising
the sets: FULLHOUSE ¼ {Set of the House variables in Table 6.6} and
FULLSENATE ¼ {Set of the Senate variables in Table 6.7}. We represent
ideology by the set of variables IDEOLOGY ¼ {ADA, NORTHEAST, SOUTH,
WEST} and interest by INTEREST ¼ {PACCORP, PACINTL, PACLABOR,
UNEMP, FOODEMP, TARIFF, NTB, NETEXP, TARIFF £ PARTY, NTB £
PARTY, NETEXP £ PARTY}. Several features are noteworthy, perhaps even
controversial, about how interest and ideology are represented. First, regional
indicators are included together with ADA to represent ideology. This represen-
tation is in line with Poole and Rosenthal (1997). While Peltzman (1985) would
disagree with the inclusion of raw ADA scores to represent ideology,
21 They are, for example, the classic Cox test, the Davidson–Mackinnon joint J-test and its extensions,
and the Mizon–Richard encompassing test.
A.K. Bohara and K. Gawande144
he would agree with region as representing ideology. We leave the task of using
appropriately corrected ADA scores (as in Levitt, 1996) to future research, using
raw ADA scores here as has been the tradition in the literature, so far. Second,
INTEREST includes three distinct sets of variables: PAC contributions
representing Olson’s (1965) special interest group theory, the constituency
variables representing Conybeare’s (1991) reduced form formulation from a
demand-supply model, and the party interaction variables motivated by the inter-
party heterogeneity hypothesis of Bailey and Brady (1998) and Irwin and Kroszner
(1999). Third, the variable PARTY is absent from both definitions. It appears only
as interactions in INTEREST. The reason we do so is to recognize the controversy
surrounding the interpretation of party. Cox and McCubbins (1993) and Peltzman
(1987) model party as mechanisms built on interest, while Poole and Rosenthal
argue that party constitutes ideology. To keep the model comparisons objective,
we present two views, one in which PARTY and INTEREST are compared
with IDEOLOGY, and another in which just INTEREST is compared with
IDEOLOGY. The marginal contribution of PARTY is thus made transparent.
Table 6.8 presents the posterior odds for comparing interest with ideology. The
left half of the table reports results from the three House models, and the right half
from the three Senate models. The results are interpreted as follows. Denote by
FULL\SETj the full model excluding the variables contained in the set SETj.
Hence, FULLHOUSE\INTEREST is the House model (as in Table 6.6) but with
the variables in the set INTEREST excluded. Then a ideology may be compared
with interest via the posterior odds ratio PðM1lyÞ=PðM2lyÞ; where M1 ¼
FULLHOUSE\INTEREST; and M2 ¼ FULLHOUSE\IDEOLOGY: A posterior
odds ratio greater than unity favors ideology, while a ratio lower than unity favors
interest.
The first row in Table 6.8 compares ideology versus interest. Consider Model 1
for the trivariate House model on protectionsim (Gephardt), the Republican
alternative bill (Michel), and passage of the House version. For the full sample of
425 observations, the ideology model is e14:7 times as likely as the interest model.
On a per observation basis, the ideology model (without interest) is 1.035 times as
likely (or 3.5% more likely) than the model of interest (without ideology). That is,
on average, ideology is 1.035 times as likely as interest to determine a trivariate
yea/nay vote for the three roll calls in Model 1. The next two rows show the
marginal contribution of adding, respectively, ideology (second row) and interest
(third row), to the baseline model consisting of committees, party, and the
constant term. The addition of the set IDEOLOGY to this baseline version of
Model 1 multiplicatively contributes e46:7 to its marginal likelihood (see Equation
2), while the set INTEREST (multiplicatively) contributes e32:0: The ratio of thesemarginal contributions is the posterior odds ratio of e14:7 in favor of ideology.
Interest and Ideology in the 1988 Omnibus Trade Act 145
Table 6.8: Model comparisons.
Theories compared House Senate
Model 1Protectionism 1
RepublicanBill 1 House Bill
Model 2GephardtRedux 1
Conference Bill
Model 3House 1
Conference 1Final Bill
Model 1Transfer ofauthority 1
Conference Bill
Model 2PresidentialDiscretion 1
Senate Bill
Model 3 Section201/301 1
Senate Bill
All obs Per obs All obs Per obs All obs Per obs All obs Per obs All obs Per obs All obs Per obs
IDEOLOGY versus
INTEREST
e14.7 1.035 e1.7 1.004 e15.7 1.038 e216.0 0.851 e27.3 0.929 e28.5 0.918
Marginal contribution:
IDEOLOGY
e46.7 1.116 e29.2 1.071 e33.1 1.081 e12.4 1.133 e23.0 1.262 e27.0 1.314
Marginal contribution:
INTEREST
e32.0 1.078 e27.5 1.067 e17.4 1.042 e28.4 1.332 e30.3 1.358 e35.5 1.431
IDEOLOGY versus
{INTEREST< PARTY}e230.2 0.931 e229.0 0.934 e214.5 0.966 e228.8 0.748 e215.9 0.852 e214.7 0.862
Notes: Variables corresponding to the models are PARTY ¼ PARTY; IDEOLOGY ¼ {Regions þ ADA}; INTEREST ¼ {PACCORP, PACINTL,
PACLABOR, UNEMP, FOOD, TARIFF, NTB, NETEXP, PARTY £ TARIFF, PARTY £ NTB, PARTY £ NETEXP}. Posterior odds based on
weak (but not flat) priors about b and V: Posterior ordinates computed using Chib’s (1995) method. Posterior odds given by (Prior Odds
Ratio) £ (Bayes Factor), where Prior Odds Ratio is set to 1, and the Bayes Factor equals the ratio of marginal likelihoods. The marginal likelihood,
where are, respectively, the data likelihood and the prior density.
A.K.Bohara
andK.Gawande
146
In per observation terms, the baseline model plus the set IDEOLOGY is 1.116
times as likely as (or 11.6% more likely than) the baseline model, while the base-
line model plus the set INTEREST is 1.078 times as likely as the baseline model.
The evidence similarly favors ideology over interest for the trivariate House
on voting on the three versions of the Omnibus bill (Model 3), but in the bivariate
model on Gephardt redux plus the conference version of the bill, interest and
ideology are nearly equally likely. At the risk of speculating, this result may be due
to the lesser contribution of the regional variables to the marginal likelihood (of the
baseline version) of Model 2 as they do in the other two models of House voting.
The posterior odds in first row of the right half of Table 6.8 show that interest
dominated ideology in all three models of Senate voting. On the issue of transfer
of authority (Model 1), the model of ideology (without interest) was e216:0 times
as likely as the model of interest. (without ideology). That is, the model of
ideology was only 0.851 times as likely as the model of interest in determining a
trivariate yes/no vote on the three roll calls, on average. Or, interest was 1.175
times as likely as (or 17.5% likelier than) interest, one average, per observation.
On the issue of Presidential discretion (Model 2), again the model of ideology was
significantly less likely than the model of interest, with ideology only 0.929 as
likely as (or 7.64% less likely than) interest in determining a trivariate vote, on
average. On the issue of protectionism and unfair trade (Model 3), again interest is
likelier than ideology.
The results require a caveat. The results as reported do not necessarily solve the
interest versus ideology debate conclusively, even within each chamber. While
our prior odds reflect objectivity, others with stronger priors in favor of ideology
may tilt the posterior odds in favor of ideology, and similarly, proponents of
interest may be able to sway the posterior odds their way. The results show,
however, that the prior odds have to be extremely strong to reverse the results
(except, perhaps, Model 2 of House voting). For example, it would take a prior
odds ratio of e14:7 in favor of interest to bring interest at par with ideology for
Model 1 of House voting, or a prior odds ratio of e16:0 against interest to bring
ideology at par with interest for Model 1 of Senate voting.
Adherents of the Cox–McCubbins or Peltzman views may yet wish to include
party as representing interest. For supporters of the parties-as-interest view, the
last row of Table 6.8 compares ideology with the model of interest but now
including party. The baseline model for these comparisons is a model with just the
committee variables and the constant term, to which are added the sets
IDEOLOGY and INTEREST < PARTY, respectively, for assessing their
marginal and relative contributions. The results for House voting now overturn
the previous finding that ideology dominated House voting (Models 1, 3) or was at
least as important as interest (Model 2). The addition of PARTY now makes the
Interest and Ideology in the 1988 Omnibus Trade Act 147
model of interest far likelier than ideology in the context of our voting data. On
protectionism and the House bill, (Model 1) interest (with party, without ideology)
is e30:2 times as likely as the model of ideology (without interest and party). That
is, interest is 1.074 times as likely as ideology in determining a trivariate yes/no
vote. The corresponding posterior odds of interest over ideology for Models 2 and
3 are, respectively, e29:0 (1.070 on a per observation basis), and e14:5 (1.035 per
observation). On the Senate side of Table 6.8, the inclusion of PARTY only
reinforces the dominance of interest over ideology as the key determinants of
voting in the Senate.
Whereas we have proclaimed agnosticism about interest versus ideology (or,
for that matter, about the prior density of the model coefficients), the purpose of a
Bayesian analysis is to confront priors with data in order to update prior beliefs.
Upon its conclusion, it requires a frank discussion of whether and how the data
analysis has updated priors. The posterior odds summarize this neatly for our
priors. Since we subscribe to the Cox–McCubbins view of parties as interest-
based mechanisms, IDEOLOGY versus INTEREST < PARTY is the relevant
comparison. Across the three house models, this more inclusive representation of
interest makes it between 3.5 and 7% likelier than ideology as a determinant of a
single multivariate observation. That does not imply ideology, per se, is
unimportant. The marginal contribution of ideology is huge, improving the
odds of the baseline model (constant, committees, party) by a remarkable 11.6%
on a per observation basis. The marginal contributions of PARTY and INTEREST
are simply greater. But an adequate explanation of voting requires not one but
both of these models.
Sponsors and supporters from both sides would find ample opportunity to take
exception to the findings. The lumping of regional variables together with ADA to
represent ideology is in discord with theories and empirical work concerning the
geographical distribution of distributive politics, for example, studies of the “1=nrule” (e.g., Weingast et al., 1981). The studies by Peltzman and Levitt cast doubt
on whether raw ADA scores genuinely represent ideology, since each has found a
significant interest component in ADA scores. The conjunction of PARTY and
INTEREST as a complete measure of interest would not go uncontested by
adherents of the ideology school, especially since the interactions of party with
constituency interests is already in the set INTEREST. The interaction address
the Baily–Brady and Irwin–Kroszner hypotheses of inter-party heterogeneity,
but do not address the Cox–McCubbins hypothesis of a mechanism designed to
sublimate the otherwise sub-optimal solution (collectively) that would result
from unbridled optimization by individual politicians. For adherents of this view,
PARTY is necessary to completely represent interest. Scholars of interest would,
in sum, find that our results understate the true odds of the model of interest.
A.K. Bohara and K. Gawande148
Wearing the ideology hat shades the picture differently. The absence of PARTY
interacted with ADA could be questioned as rightfully belonging in the set of
variables of ideology. Similarly, regional variables should be interacted with
PARTY. We could not because, first, no satisfactory theory about ADA £
PARTY has been put forth, and second, interacting PARTY with the regional
variables causes a problems in the estimation due to collinearity.22 We leave the
variety of extensions this discussion opens up, to future research.
For the sake of argument, one we think would attract a large number of
participants, suppose the first row of Table 6.8 were an accurate depiction of the
interest versus ideology debate. Then it is plain to see that ideology mattered more
than interest to House voting, while in the Senate it was driven more by interest,
than ideological voting. If the INTEREST indeed completely represents interest,
than that is what our study finds to be a satisfactory determination. It also begs the
question of why such inter-chamber heterogeneity must exist. In closing, we
examine two possible answers.
Sinclair (1992) advances stronger House leadership as a possible reason for
the inter-chamber difference. She presents evidence demonstrating ideological
heterogeneity, which nearly split the Democratic party in 1970s, decreased in the
1980s as big deficits shrank the issue space. Deep policy divisions arose between
the President and House Democrats 1980s. These changes, Sinclair argues,
lowered costs and increased benefits of leadership in the House, from which
emerged strong leadership in the 1980s in place of the legislative entrepreneurship
that characterized House Democrats in the previous decade. Strong leadership
made it easier for Democrats to form an ideologically homogeneous alignment,
especially on issues that pitted them against the President. Our own reading of the
evidence from floor action on the House and Senate bills (Congressional
Quarterly, 1987–1988; Ehrenhalt et al., 1987; Schwab, 1994) is that individual
entrepreneurship characterized action on the Senate but not House floor. There
were nearly 160 amendments offered on the Senate floor, many resolved through a
voice vote, but only a handful offered on the House floor. The strong finding of
both PARTY and ADA in the results reported earlier can both be explained by
Sinclair’s theory.
A second explanation may be found in differential voting in systems of low
versus high “party discipline”. McGillivray (1997) hypothesizes that in high-
discipline systems (e.g., majoritarian systems like Canada, UK, France)
22 As it is, the number of variables is quite large relative to the sample size. The inclusion of even 18
variables in the full model causes the estimate of the marginal likelihood (but not the coefficient
estimates, which are quite stable) to be sensitive to small changes in parameter estimates. The marginal
likelihoods for the smaller models upon which the model comparisons are based are quite stable.
Interest and Ideology in the 1988 Omnibus Trade Act 149
representatives toe the party line. In low party discipline systems (e.g., US)
legislators may vote for constituency preferences over party preferences.
An extension of a discipline-based answer to voting on similar issues by the
two chambers may provide possibly revealing answers. Empirically, this would
require a clear measure of what constitutes low discipline. McGillivray (see also
Conybeare) uses district marginality as measures of the level of discipline.
6.8. CONCLUSION
This chapter is motivated by the need to resolve more than one controversy about
the determinants of voting, and hence is wide in scope. The setting is the
legislation of the most important trade act in recent US history, the 1988 Omnibus
Trade and Competitiveness Act. Hence it investigates the political economy of
trade policy at its grassroots, which filters eventually into tangible redistributive
outcomes across sectors. The main focus of the chapter is upon formally
distinguishing the contribution of ideology and interest towards explaining voting
behavior. This is a richly debated but as yet unresolved question in the literature.
Since the analysis is over a set of interrelated roll call votes on the common,
binding trade theme the appropriate methodology is a MVP model that takes
account of a correlations across roll call votes. Estimation and model comparisons
are performed in the context of the MVP model using the recently developed
computational Bayesian methodology of Chib and Greenberg (1998). No stronger
than uninformative priors are used, so that the results are based largely on data
evidence, not our prior opinions. We believe this to be the first formal comparison
of ideology versus interest in the literature.
Our results should be of interest, methodologically as well as substantively.
The results show that an informal comparison of models based on size, sign, and
statistical significance of estimates, as is the state of the art in the literature, is
an inadequate basis for discriminating between models. Tables 6.6 and 6.7 show
all theories to be effective in explaining at least some of the voting, without being
able to reject any theory outright. Yet the model comparisons favor interest as
the relatively more important contributor to explaining Senate voting, but show
that the reverse is true for House voting if PARTY is taken to be a neutral
variable (if it is lumped with interest, then interest dominates ideology in House
voting as well). The model comparisons with baseline models show also that
individually, the contribution of ideology and interest are both substantial. The
results highlight inter-chamber differences in voting, which beg explanation. We
offer one based on Sinclair’s finding that effective leadership promoted more
A.K. Bohara and K. Gawande150
ideological homogeneity in the House during that period, while individual
legislative entrepreneurship characterized voting in the Senate.
In closing, it is appropriate to ask what might comprise “ideology” in trade
voting. Issue-specific ideology on trade issues might be very different from
general liberal-conservative ideology. Bohara et al. (2003) address this issue using
factor analytic methods, and describe the nature of trade-specific ideology. They
find that trade-specific ideology, as different from general ideology, lay at the
heart of voting on the 1988 Omnibus Act. More research on those lines would be
helpful contributions in the ideology versus interest debate.
ACKNOWLEDGEMENTS
Financial support from UNM RAC and TAC grants is acknowledged. We are
grateful to Stanley Nollen for providing PAC contribution and other data, and Sid
Chib for generosity with his code and comments. All remaining errors are ours.
APPENDIX A. BAYESIAN ANALYSIS OF THE MULTIVARIATE
PROBIT MODEL
The description of Bayesian analysis of the MVP model borrows from Chib and
Greenberg (1998). Consider the J-variate probit model where Yij is a random
variable denoting a binary 0/1 response on the ith observation on the jth variate.
Let yi ¼ ðyi1;…; yiJÞ0; 1 # i # n; J-variate realization on the ith observation. The
MVP model is conditioned on the parameters b : k £ 1; and S : k £ k and a set of
explanatory variables Xi; as follows:
yi ¼ Xibþ ei; ðA1Þ
where the data yi : j £ 1 as described above, Xi is a ðJ £ kÞ matrix of explanatory
variables given as diagðxi10;…; xiJ
0Þ; where xij : kj £ 1 is the set of kj explanatory
variables in the jth equation (hence k ¼P
Jj¼1 kjÞ; b : k £ 1 is the vectors of
coefficients on the explanatory variables, and ei : J £ 1 is the (correlated) vector of
residuals with zero mean and correlation matrix S : J £ J: In the Bayesian
analysis of the MVP model, b is a random vector, which is assigned a prior
distribution. Bayesian inferences proceeds with the analysis of the posterior
distribution, which combines (actually updates) the prior pdf with the data
likelihood. In order to proceed with the likelihood function for the MVP model,
define the J-variate normally distributed latent variable Zi ¼ ðzi1;…; ziJÞ with
Interest and Ideology in the 1988 Omnibus Trade Act 151
distribution Zi , NJðXib;SÞ: Now let the realization yij equal 1 if zij . 0 and 0
otherwise:
yij ¼ Iðzij . 0Þ; j ¼ 1;…; J; ðA2Þ
where IðAÞ is the indicator function of the event A: The joint density of the data,
conditioned on b; S; and Xi; is given by
prðyilb;S;XiÞ ¼ð
Bij
· · ·ð
Bij
fJðZilXib;SÞdZi; ðA3Þ
where Bij; is the positive interval ð0;1Þ if yij ¼ 1 and the nonpositive interval
ð21; 0� if yij ¼ 0; and fJð·lXib;SÞ is the J-variate normal density with mean
vector Xib; and correlation matrix. Let Bi ¼ B11 £ B12 £ · · · £ B1J and note that
the set Bi depends only on the value of yi and not the parameters. Finally, denote
the p ¼ JðJ 2 1Þ=2 free parameters of S by s ¼ ðs12;s13;…;sJ21;JÞ:Suppose we have a random sample of nJ observations y ¼ ð y1;…; ynÞ: The
posterior density of the parameters of the MVP model, by Bayes theorem, is
given by
pðb;s; ZlyÞ / pðb;sÞf ðZlb;SÞprðylZ;b;SÞ; b [ RK ; s [ C; ðA4Þ
where Z ¼ ðZ1;…; ZnÞ; and pðb;sÞ is a joint prior density on the parameters
ðb;sÞ; and C is a convex set contained in the hypercube ½21; 1�p since the
elements of s form a correlation matrix. From the mapping in Equation A2 it is
clear that prðylZ;b;SÞ ¼ IðZi [ BiÞ: Hence the posterior probability in Equation
A4 can be written as
pðb;s; ZlyÞ / pðb;sÞYn
i¼1
fJðZilb;SÞIðZi [ BiÞ; ðA5Þ
where
fJðZilb;SÞ / lSl21=2
exp½2 12ðZi 2 XibÞ
0S21ðZi 2 XibÞ�Iðs [ CÞ:
Hence evaluation of the data likelihood (of y) is not required since conditioning on
Bi determines the 1/0 probability of yi:The estimation of the posterior mean of the parameters as well as the
posterior ordinate at the posterior means (for model comparisons) is done via
A.K. Bohara and K. Gawande152
the powerful computational methods involving MCMC sampling. Even if the
joint density (posterior density or data likelihood) is not analytically well
defined, it can be broken down into the product of conditional densities that are
analytically well defined and can be used to sample synthetic data. The Gibbs
sampler derives its importance from the fact that generating sequences of
samples from the conditional distributions upon updating the conditioning
parameters results is a sample that is actually generated from the joint density
itself. The generated sequence can then be used to compute the posterior mean
(or ML estimate) and also the ordinate of the posterior density (or value of the
likelihood function).
In the MVP model the Markov chain sampling scheme is constructed from
the conditional distributions in Equation A4: ½Zilyi;b;S� for i # n; ½bly; Z;S�and ½sly; Z;b�: Then a Gibbs sequence {ZðjÞ;bðjÞ;sðjÞ} can be generated by the
following algorithm:
1. Specify starting values, say ML estimates, ðbð0Þ;sð0ÞÞ; and set i to 0.23
2. Simulate Zðiþ1Þ from ½Zly;bð0Þ;Sð0Þ�;
Simulate bðiþ1Þ from ½bly; Zðiþ1Þ;Sð0Þ�;24 and
Simulate sðiþ1Þ from ½sly; Zðiþ1Þ;bðiþ1Þ�;25
3. Set i ¼ iþ 1 and go to step 2.
A Markov chain {Zð0Þ;bð0Þ;sð0Þ}; {Zð1Þ;bð1Þ;sð1Þ}; {Zð2Þ;bð2Þ;sð2Þ};… is thus
produced where each subsequent item of the chain is simulated using the full
conditional densities where the conditioning elements are revised during an
iteration. After discarding the first l realizations of the sequence (where l varies
depending upon the application), the next M realizations can be used for
computing the characteristics of the marginal distributions of b and s (for
example, the posterior means are simply the sample mean of the generated bs andssÞ; or approximating the marginal densities themselves.
The approximation of marginal densities is an attractive feature of the MCMC
method, for it allows the comparisons of competing models on the basis of
posterior odds or odds based on likelihoods. Specifically, model comparisons
require the computation of marginal data densities. Consider a set of K models,
23 One can set S to be diagonal and estimate the J equations separately as single equation probits in
order to supply initial values of s:24 The sampling of b from ½bly;Z;S�; is straightforward, from a multivariate normal distribution as in
Chib and Greenberg (1998, (6)).25 The careful sampling of s is done via the Metropolis–Hastings algorithm (Chib and Greenberg
1995). This is an innovation in the Chib and Greenberg (1998) paper and is described there.
Interest and Ideology in the 1988 Omnibus Trade Act 153
indexed as Mk; k ¼ 1;…;K: The marginal likelihood of Mk is defined as
mðylMkÞ ¼ð
prðylMk;b;SÞpðb;slMkÞdb ds; ðA6Þ
where prðylb;s;MkÞ; pðb;slMkÞ are, respectively, the data likelihood and the
prior density for model Mk: The integration in Equation A6 produces a weighted
average of the data density where the prior density provides the weights. However,
the integration in Equation A6 is analytically impossible for a large class
likelihoods, certainly the MVP models analyzed here. It is also computationally
nontrivial in general. Chib (1995) provides a simple way of computing the
marginal density as a by-product of the Gibbs sampling output. Chib’s idea is to
rewrite Bayes formula for the posterior parameter density (suppressing the index
for model k)
pðb;slMk; yÞ ¼prðylMk;b;SÞpðb;slMkÞ
mðylMkÞðA7Þ
as
mðylMkÞ ¼prðylMk;b;SÞpðb;slMkÞ
pðb;slMk; yÞ: ðA8Þ
Equation A8 is an identity in ðb;sÞ and so may be evaluated at any point.
Following Chib (1995) a high-density point such as the posterior mean ðbp;spÞ is
used. Using the conditional probability rule and taking logs
log mðylMkÞ ¼ log prðylMk;bp;Sp
Þ þ log pðbplMkÞ þ log pðsp
lMkÞ
2 log p ðbplMk; y;s
pÞ2 log pðsp
lMk; yÞ; ðA9Þ
where
fðs2plyÞ ¼
1
M
XM
g¼1
f ðs2ply; ypðgÞ;bðgÞ
Þ; ðA10Þ
and
fðbply;s2p
Þ ¼1
M
XM
g¼1
f ðbply; ypðgÞ;s2p
Þ: ðA11Þ
The numerator in Equation A8 is computed from the first three terms on the
RHS of Equation A9 evaluated at the posterior parameter means, bp and s2p; from
A.K. Bohara and K. Gawande154
the M Gibbs realizations. The denominator in Equation A8 is evaluated as the
mean pdf over the M Gibbs realizations. Since we are sampling over three vector
blocks ðb;s2; ypÞ; the evaluation of the denominator is done in two steps. The first
step consists of generatingM Gibbs realizations from the full conditional pdfs and
computing the posterior ordinate fðs2plyÞ at s2p as described in Equation A10. The
second step consists of generating a new set of M Gibbs realizations from the
reduced conditional pdfs, where the conditioning is on the value s2p computed
from the first M realizations, and computing the posterior ordinate ln fðbply;s2pÞ
at bp (from the second run) and s2p:The posterior odds of M1 relative to M2 is given by the product of the ratio of
their marginal data densities, called the Bayes factor, with the prior odds ratio
where the prior odds ratio is specified by the researcher,
PðM1lyÞ
PðM2lyÞ¼
f ðylM1Þ
f ðylM2Þ
PðM1Þ
PðM2Þ: ðA12Þ
The numerator and denominator of the Bayes factor are computed, respectively,
from the Gibbs output for each model by exponentiating Equation A9. The
posterior odds thus computed are the basis for the model comparisons performed
in this chapter. We use weak priors so that the results are based largely on infor-
mation in the data.
REFERENCES
Adler, E. S. and Lapinski, J. S. (1997). Demand side theory and congressional committee
composition: a constituency characteristics approach. American Journal of Political
Science, 41, 895–918.
Bailey, M. (2001). Quiet influence: the representation of diffuse interests, on postwar trade
policy. Legislative Studies Quarterly, 26, 1–36.
Bailey, M. and Brady, D. W. (1998). Heterogeneity and representation: the senate and free
trade. American Journal of Political Science, 42, 524–544.
Baldwin, R. E. (1986). The Political Economy of U.S. Import Policy, Cambridge, MA: MIT
Press.
Bhagwati, J. N. and Patrick, H. T. (eds.) (1990). Aggressive Unilateralism: America’s 301
Trade Policy and the World Trading System, Ann Arbor, MI: University of Michigan
Press.
Bohara, A. K., Camargo, A. I., Grijalva, T., Gawande, K. (2004). Fundamental dimensions
of U.S. trade policy. Journal of International Economics, in preparation.
Chib, S. (1995). Marginal likelihood from the Gibbs output. Journal of the American
Statistical Association, 90, 1313–1321.
Interest and Ideology in the 1988 Omnibus Trade Act 155
Chib, S. and Greenberg, E. (1995). Understanding the Metropolis–Hastings algorithm.
American Statistician, 49, 327–335.
Chib, S. and Greenberg, E. (1996). Markov Chain Monte Carlo simulation methods in
econometrics. Econometric Theory, 12, 409–431.
Chib, S. and Greenberg, E. (1998). Analysis of multivariate Probit models. Biometrika, 85,
347–361.
Cohen, R. E. and Schneider, W. (1989). Congressional report: slim pickings. National
Journal, 203–233.
Congressional Quarterly. Various issues in 1987 and 1988.
Conybeare, J. A. C. (1991). Voting for protection: an electoral model of tariff policy.
International Organization, 45, 57–81.
Cox, G. W. and McCubbins, M. D. (1993). Legislative Leviathans, Berkeley: University of
California Press.
Ehrenhalt, A., Amrine, R. and Duncan, P. D. (eds.) (1987). 1987/88 Politics in America:
The 100th Congress, Washington, DC: CQ Press.
Greene, W. H. (1993). Econometric Analysis, 3rd ed., New Jersey, Prentice Hall.
Heckman, J. and Snyder, J. (1997). Linear probability models of the demand for attributes
with an empirical application to estimating the preferences of legislators. Rand Journal
of Economics, 28, S142–S189.
Irwin, D. A. (1994). The political economy of free trade: voting in the British general
election of 1906. Journal of Law and Economics, 37, 75–108.
Irwin, D. A. and Kroszner, R. S. (1999). Interests, institutions, and ideology in securing
policy change: the Republican conversion to trade liberalization after Smoot–Hawley.
Journal of Law and Economics, 42, 643–673.
Kalt, J. P. and Zupan, M. A. (1984). Capture and ideology in the economic theory of
politics. American Economic Review, 74, 279–300.
Kau, J. B. and Rubin, P. H. (1979). Self-interest, ideology and logrolling in congressional
voting. Journal of Law and Economics, 22, 65–84.
Kau, J. B., Keenan, D. and Rubin, P. H. (1982). A general equilibrium model of congres-
sional voting. Quarterly Journal of Economics, 97, 271–293.
Krugman, P. M. (1986). Strategic Trade Policy and the New International Economics,
Cambridge: MIT Press.
Levitt, S. D. (1996). How do denators vote? Disentangling the role of voter preferences,
party affiliation, and senator ideology. American Economic Review, 425–441.
Levitt, S. D. and Snyder, J. M. (1995). Political parties and the distribution of federal
outlays. American Journal of Political Science, 39, 958–980.
McGillivray, F. (1997). Party discipline as a determinant of the endogenous formation of
tariffs. American Journal of Political Science, 41, 584–687.
Nollen, S. and Quinn, D. P. (1994). Free trade, fair trade, strategic trade and protectionism
in the U.S. congress, 1987–88. International Organization, 48, 491–525.
Olson, M. (1965). The Logic of Collective Action, Cambridge: Harvard University Press.
Peltzman, S. (1984). Constituent interest and congressional voting. Journal of Law and
Economics, 27 (April), 181–210.
A.K. Bohara and K. Gawande156
Peltzman, S. (1985). An economic interpretation of the history of congressional voting in
the twentieth century. American Economic Review, 75, 656–675.
Poole, K. T. and Rosenthal, H. (1997). Congress: A Political Economic History of Roll Call
Voting, Oxford: Oxford University Press.
Schickler, E. and Rich, A. (1997). Controlling the floor: parties as procedural coalitions in
the House. American Journal of Political Science, 41, 1340–1375.
Schwab, S. (1994). Trade-Offs: Negotiating the Omnibus Trade and Competitiveness Act,
Boston: Harvard Business School Press.
Shepsle, K. A. and Weingast, B. R. (1987). The institutional foundations of committee
power. American Political Science Review, 81, 85–104.
Sinclair, B. (1992). The emergence of strong leadership in the 1980s house of
representatives. Journal of Politics, 54, 657–684.
Stratmann, T. (1992). The effects of logrolling on congressional voting. American
Economic Review, 82, 1162–1176.
Tosini, S. C. and Tower, E. (1987). The textile bill of 1985: the determinants of
congressional voting patterns. Public Choice, 54, 19–25.
Trefler, D. (1993). Trade liberalization and the theory of endogenous protection: an
econometric study of U.S. import policy. Journal of Political Economy, 101, 138–160.
US Department of Commerce (1984). Bureau of the Census, Congressional District Atlas
(98th Congress).
Weingast, B. R. and Marshall, W. J. (1988). The industrial organization of congress; or why
legislatures, like firms, are not organized as markets. Journal of Political Economy, 96,
132–163.
Weingast, B. R., Shepsle, K. A. and Johnson, C. (1981). The political economy of benefits
and costs: a neoclassical approach to distributive politics. Journal of Political Economy,
89, 642–664.
Zellner, A. (1971). An Introduction to Bayesian Inference in Econometrics, New York:
John Wiley.
Interest and Ideology in the 1988 Omnibus Trade Act 157
CHAPTER 7
Industry and Factor Linkages
Between Lobby Groups
CHRISTOPHER MAGEE
Department of Economics, Bucknell University, Lewisburg, PA 17870, USA
Abstract
Factor mobility implies that individuals owning a common factor of production
have more in common than individuals owning factors within the same industry.
A simple test of factor mobility, then, involves determining whether a political
action committee’s (PAC) pattern of campaign contributions is more similar to the
patterns of other groups representing the same factor or of groups within the same
industry. Using data on campaign contributions during the 1991–1992 election
cycle, this chapter examines the correlation between different PACs’ contributions
to candidates for the House of Representatives. The results have broad
implications for models of factor mobility and lobbying behavior.
Keywords: Lobbying, contributions, factor mobility
JEL classification: D7
7.1. INTRODUCTION
The traditional view that governments attempt to choose policies solely for the
good of society has been overturned in recent years after persistent evidence to
the contrary. Trade barriers, tax loopholes, and agricultural subsidies are merely a
E-mail address: [email protected]
few of the cases in which public policies contradict the notion that governments
act to maximize social welfare. As one of the characters in Alfred Hitchcock’s The
Lady Vanishes says: “The theory was fine. The facts were misleading.” Political
economy research on policy formation has usually emphasized the important role
that pressure politics by interest groups plays in moving the government away
from welfare-maximizing decisions.
Grossman and Helpman (1994) is a prime example. In their model, a policy
maker values both social welfare and campaign contributions from interest
groups. If an import-competing industry is represented by an organized lobby
group, it offers contributions to the policy maker in exchange for tariff protection
(export subsidies for an exporting industry). The policy maker balances his or her
gains from receiving interest group money against the losses the tariff creates for
society. Without any organized interest groups, the welfare-maximizing trade
policy (free trade) would be chosen. With some (but not all) members in society
represented by interest groups, however, the endogenous policy leads to an
inefficient outcome for society. Thus, interest group lobbying is the critical factor
leading to poor economic policies. A similar conclusion can be drawn from other
models of policy formation, such as Magee et al. (1989).
As Olson (1965) discusses, however, it is very difficult for large groups to
overcome collective action problems and organize an effective lobby. The
national lobbies that exist are often federations of smaller groups, which are more
capable of overcoming free-rider problems. Labor unions, for example, form at
the local level and then group together nationally in many of their lobbying efforts.
This coordination between interest groups that initially form separately is a
characteristic of many lobbying efforts. In attempts to get the NAFTA and GATT
Uruguay Round trade deals approved in the US Congress, for instance, corporate
interest groups representing many different industries joined together in putting
pressure on legislators. Many different labor groups, on the other hand, joined
together to oppose approval of these trade negotiations.
The nature of interest groups joining together in common lobbying efforts can
reveal a good deal about the preferences of the groups. Magee (1980) shows that
examining whether different interest groups lobby on the same or opposite sides of
a trade bill provides a simple test of the Stolper–Samuelson (S–S) theorem. This
theorem predicts that all members of one factor of production (the abundant
factor) will gain from trade liberalization, while all members of the scarce factor
of production in a country will lose from expanded trade. In the 2 £ 2 Heckscher–
Ohlin model, then, we should observe capital and labor groups lobbying on
opposite sides of trade bills. If, on the other hand, factors are specific to their
industry, then trade liberalization benefits those factors in net exporting industries
and hurts factors in import-competing industries. The prediction from this model
C. Magee160
is that both capital and labor groups within an industry will lobby on the same side
of a trade liberalizing bill—both supporting it, if their industry is an exporting
industry and both opposing trade liberalization, if the industry is import-
competing. Magee found scant support for the S–S theorem, as capital and labor
groups adopted the same position on the Trade Act of 1974 in 19 out of 21
industries examined.
A recent paper by Beaulieu and Magee (2004) performs a similar test for
lobbying over the NAFTA and GATT Uruguay Round bills and finds stronger
support for the S–S prediction than that in Magee (1980). Consistent with the S–S
theory, the paper shows that capital groups gave much larger shares of their
contributions to politicians supporting the NAFTA and GATT bills than did labor
unions. There was very little evidence, particularly for capital groups, however,
that net exporting industries were more likely to support trade liberalization than
import-competing industries. The contribution patterns thus provide indirect
support for the S–S assumption of factor mobility over the specific factors
assumption of factor immobility.
One explanation for the different results in Magee (1980) and Beaulieu and
Magee (2004) is the time horizon of the bills being examined. Magee (1980)
examined the Trade Act of 1974, a bill that was likely to be modified within 5 or 6
years. Thus, short-run interests are likely to dominate lobbying efforts over the
bill. Beaulieu and Magee (2004) examine lobbying over the NAFTA and GATT
agreements, which are unlikely to be overturned in the foreseeable future, so that
long-run considerations are paramount in determining interest groups’ lobbying
positions. The specific factors model is a short-run model of the economy while
the Heckscher–Ohlin model (with its assumption of perfect factor mobility) is a
better description of the economy in the long run. Thus, it makes sense that
lobbying over the easily modified 1974 Act would be consistent with the
predictions of the specific factors model, while lobbying over the longer run
NAFTA and GATT agreements would be consistent with the Heckscher–Ohlin
predictions.
A paper by Hartigan and Tower (1982) reveals a trade policy reason why factor
mobility is an important issue. Hartigan and Tower find that when factors are
mobile, US tariffs and quotas have a much larger impact on real income and its
distribution than when labor is assumed to be the only mobile factor. Combined
with this result, then, the Beaulieu and Magee (2004) evidence in favor of factor
mobility suggests that trade barriers may be an important determinant of US
incomes and inequality.
This chapter expands on the idea that the relationship between different interest
groups’ contribution patterns can reveal important information about their
preferences. Although the Beaulieu and Magee (2004) investigation was specific
Industry and Factor Linkages Between Lobby Groups 161
to trade policy, the point is a general one—lobby groups with common interests
will have an incentive to contribute to the same set of politicians. Thus, if labor
groups have a common interest in promoting certain candidates over others, we
should observe a strong correlation between the contribution patterns of different
labor PACs. Such a common interest would arise if labor is very mobile across
sectors. If industry interests are paramount in determining interest group
preferences over public policy, on the other hand, we should observe a large
correlation between contribution patterns of different PACs representing the same
industry. Such a result would support the view that factors earn sector-specific
rents and thus have an incentive to promote the fortunes of their industry. This
chapter examines the relationships between interest group contribution patterns in
order to investigate whether common lobbying efforts occur within industries or
within factors of production.
The contrast between the results in Magee (1980) and Beaulieu and Magee
(2004) suggests that industry and factor will both play roles in determining the
relationship between interest groups’ contribution patterns. Short-run consider-
ations give political action committees (PACs) representing the same industry an
incentive to give money to a common set of politicians, while long-run
considerations suggest that representing the same factor of production leads to
similar contribution patterns. Because interest groups have both short and long-
run goals in their lobbying efforts, both industry and factor should play a role in
determining the relationship between interest groups’ contributions across
candidates.
Section 7.2 presents a simple model revealing the incentives that PACs have to
donate money to candidates for political office. The model shows that under
certain general conditions, lobby groups that have common policy interests will
tend to concentrate their contributions on the same set of candidates for office.
Section 7.3 presents empirical tests of the relationship between interest group
contribution patterns. The final section concludes.
7.2. THEORY
This section presents a simple theoretical model of PAC contribution decisions. It
reveals that rational interest groups give money to candidates for three primary
reasons. First, they may wish to influence the outcome of elections, with the desire
of helping more favorable candidates gain office. Second, the interest groups may
hope to gain direct influence over the policy decisions made by candidates who
have won election to the legislature. Finally, PACs may be seeking to gain
unobserved services that legislators can provide. Grossman and Helpman (1996)
C. Magee162
refer to the first motivation for PAC contributions as the electoral motive and the
second as the influence motive. Hall andWayman (1990) present evidence that the
unobserved service motivation is also important. They show that contributions
from PACs lead representatives to spend more time and effort in committees
dealing with issues that are important to the interest group.
For the sake of concreteness, let us examine an interest group that is interested
in only one policy initiative being considered in the upcoming legislative
session—a proposal to liberalize trade with foreign countries. Consider an interest
group ðiÞ that receives an expected present discounted value of lifetime utility ViL
if the liberalization is approved by the legislature and Við2LÞ if the measure is
defeated. The group’s expected income net of lobbying costs is Ri ¼
gViL þ ð12 gÞVið2LÞ 2P
j cij; where g is the probability that the trade liberal-
ization package is approved, and cij represents interest group i’s contribution to
candidate j: The group chooses its contribution levels to maximize expected
income, so it sets
›Ri
›cij¼
›g
›cijðViL 2 Við2LÞÞ2 1 ¼ 0; when cij . 0: ð1Þ
For politicians who do not receive any campaign money from the group ðcij ¼ 0Þ;it must be the case that
›Ri
›cij
�����cij¼0
¼›g
›cijðViL 2 Við2LÞÞ2 1 , 0:
Assume for the moment that candidates choose their policy positions in
advance of the election and that interest groups then give money to help their
favored candidates win election, as in Magee et al. (1989). Let u be the probabilitythat any one vote in the legislature is decisive in determining the outcome of
the policy initiative. The effect of a contribution on the likelihood that the bill
passes is
›g
›cij¼ u
›pj
›cijðLj 2 L2jÞ; ð2Þ
where pj is the probability candidate jwins election, Lj is the probability candidate
j votes for the liberalization initiative, and L2j is the probability that candidate j’s
opponent would vote for the bill. Substituting Equation 2 into Equation 1 reveals
Industry and Factor Linkages Between Lobby Groups 163
that an interest group gives money until
ðLj 2 L2jÞðViL 2 Við2LÞÞ ¼1
u›pj
›cij
: ð3Þ
Assuming that contributions to a candidate raises his or her probability of
winning the election, the right-hand side of Equation 3 is positive. Thus, it must be
the case that ðViL 2 Við2LÞÞ and ðLj 2 L2jÞ are the same sign. Interest groups
favoring trade liberalization will give money to a candidate only if he has a larger
probability of voting for freer trade than his opponent. Interest groups opposing
free trade will support candidates for office who are more likely to vote against
trade liberalization. A common assumption (made for example, by Magee et al.
(1989)) is that contributions received by a candidate have a diminishing marginal
effect on the election outcome. In terms of Equation 3, this assumption means that
›pj=›cij falls, and the right-hand side of the equation rises, as cij increases. With
candidates having predetermined policy positions, then, Equation 3 shows that cijrises with lViL 2 Við2LÞl: Interest groups that have stronger preferences for or
against trade liberalization will give more money to candidates who, respectively,
support or oppose freer trade. This result is quite intuitive—PACs have greater
incentive to help a candidate win election when the candidate strongly supports
the interest group’s favored position and when her opponent strongly disagrees
with the PAC.
The model thus shows that interest group preferences over trade policy are
revealed through their campaign contributions if the groups are giving financial
contributions in order to help favored candidates win election. The model can
easily be expanded to consider influence and service motivations for donating
campaign funds. Consider an interest group whose net income is
Ri ¼X
j
Sij þ gViL þ ð12 gÞVið2LÞ 2
X
j
cij; ð4Þ
where Sij represents the services provided by legislator j to PAC i: These servicescan be thought of as including all non-trade motivations for a PAC to donate
money to a campaign. Suppose that political candidates have some likelihood of
supporting trade liberalization, but that they are amenable to persuasion, so that
campaign dollars allow the interest group to gain access to the politician and alter
the probability he or she supports the trade bill. The effect of contributions on
C. Magee164
the probability the liberalization measure passes the legislature is
›g
›cij¼ u
›pj
›cijðLj 2 L2jÞ þ pj
›Lj
›cij
" #
: ð5Þ
Differentiating Equation 4 and substituting Equation 5 into the result shows that
PAC i gives money to candidate j until
›pj
›cijðLj 2 L2jÞ þ pj
›Lj
›cij¼
12›Sij
›cij
uVi
; when cij . 0; ð6Þ
where Vi ¼ ViL 2 Við2LÞ is the interest group’s net gain from liberalization. If the
interest group values the electoral and influence gains from contributing, it must
be the case that the marginal gain to the interest group in terms of services
provided by the candidate is less than one. The numerator in the right-hand side of
Equation 6 is thus non-negative, and it is strictly positive for PACs that hope to
affect election outcomes or to influence policy choices through their contributions.
The right-hand side of the equation, then, is positive for interest groups favoring
liberalization ðVi . 0Þ and negative for those opposing it.
The first term in Equation 6 represents the benefit to the lobby of helping a free
trade or protectionist candidate win election. The second term shows the gain from
changing the probability that candidate j supports the liberalization bill proposed.
These two terms correspond to the electoral and influence motives discussed
earlier for campaign contributions. In most models of probabilities, such as the
probit model, the effect of campaign contributions on the likelihood of a candidate
supporting a bill ð›Lj=›cijÞ is maximized when the candidate has a 50%
probability of supporting the trade deal. Thus, the influence motive pushes interest
groups to give money to candidates who are undecided. The electoral motive, on
the other hand, induces PACs favoring liberalization to give money to candidates
who are already strong supporters of free trade (meaning that there is a large
Lj 2 L2j in Equation 6). As long as the electoral motive is a consideration in its
contribution decisions, PACs favoring liberalization will concentrate their money
on candidates who are inclined to support free trade on average while lobby
groups favoring trade barriers will give money to candidates who are more
protectionist. Magee (2002) finds empirical evidence that PACs are motivated by
the desire to affect elections. On four out of five issues examined, including trade
policy, PAC contributions flowed more readily to 1996 candidates for the House
of Representatives who had already committed to supporting the interest group’s
Industry and Factor Linkages Between Lobby Groups 165
preferred policy stance. Thus PACs will, in general, contribute more to candidates
who are likely to vote in the group’s interest.
The model shows that, with an electoral motive for giving contributions,
two different PACs both of which favor trade liberalization have incentives to
give contributions to the same group of candidates for office—those strongly
favoring free trade and whose opponents are protectionists. PACs favoring
protectionism have an incentive to give to a different group of candidates,
with all such protectionist interest groups concentrating their donations on
those candidates favoring trade barriers.
If interest groups give money primarily for reasons of gaining influence
with legislators, the prediction about patterns of PAC contributions is slightly
different. In this case, both PACs favoring trade liberalization and those
favoring protectionism will tend to give to the same group of candidates—
those with a high probability of winning election but who are undecided on
the trade liberalization issue. Thus, all interest groups who are concerned with
trade bills, whether protectionist or not, will give money to a similar group of
candidates. PACs whose interests lie in other areas of legislation, on the other
hand, will not necessarily concentrate their contributions on this same group
of candidates.
Finally, consider the case in which interest groups give money primarily to gain
unobservable services that legislators can provide. The ability for a legislator to
provide services to interest groups depends on holding positions of power in
Congress or membership on committees that deal with the policies the interest
group is concerned with. Thus, all PACs have an incentive to give contributions to
representatives who have achieved positions of power. PACs with common
concerns will tend to concentrate their donations on politicians who are members
of relevant committees. As with the influence motive, side of a policy debate
which the interest group favors is irrelevant in considering the relationship
between its contribution pattern across candidates and that of other interest groups.
Whether interest groups are motivated by electoral, influence, or service
considerations in giving campaign contributions, we should observe some positive
correlation between the contribution patterns for different PACs across candidates
since all PACs will concentrate their donations on those politicians most able to
provide services to interest groups. A larger correlation between two PACs’
contribution patterns will be evident when the groups are interested in the same
policy issues. The strongest correlation between two PACs’ contribution patterns
will occur when both interest groups are concerned with the same issues and have
the same preferences over the policies adopted. Examining how similar one
interest group’s contributions are to another group’s donations reveals the extent
to which the two groups have common interests in promoting certain policies.
C. Magee166
The following sections of the chapter examine the relationships between
different interest groups’ contribution patterns across candidates. The question
addressed is whether interest groups representing a common factor of production
have more similar contribution patterns than interest groups representing a
common industry. If corporate PACs are primarily interested in certain issues
while labor PACs are interested in other issues, for example, there will be a
stronger correlation between the contribution patterns of two corporate PACs (or
two labor PACs) than there will be between the contribution patterns of a
corporate and a labor PAC. If PACs are divided along industry lines, however,
there should be a significant positive correlation between the contribution
decisions of interest groups within the same industry.
Examining the relationships between contribution decisions of different
interest groups, then, reveals whether groups lobby along factor or industry
lines. This question is relevant for the debate over how mobile factors are across
industries and whether short or long-run goals are more important to PACs in
their lobbying efforts.
7.3. EMPIRICAL ANALYSIS
The data on PAC contributions were provided by the Federal Election
Commission. The data set includes only qualified and non-qualified non-party
PACs who gave money to at least one candidate for Congress in 1991–1992.
Independent expenditures against a candidate and communications costs against a
candidate are counted as negative contributions. There are 482 manufacturing
PACs whose industry and factor are identified, and they gave to 1988 candidates
for the House of Representatives and the Senate. Of these interest groups, 435 are
corporate interest groups and 47 are labor groups. The disparity in numbers
between corporate and labor groups reflects in part the difficulty of organizing for
labor unions relative to firms. Since many industries are dominated by a few large
firms, it is easier for them to overcome the collective action problems that plague
lobby groups. Unions, on the other hand, must successfully persuade many
different workers to sacrifice personally in pursuit of a collective good. As Olson
(1965) points out, the difficulties inherent in organizing a large group mean that
businesses will be better organized than labor. The small number of labor unions is
also a result of having a few large manufacturing labor groups rather than many
small ones.
Table 7.1 presents the correlations between the contribution patterns of PACs
for every pair of PACs in the data set. With 482 PACs, there are 115,921 pairs of
PACs that can be examined. Of these pairs of interest groups, 18,295 pairs
Industry and Factor Linkages Between Lobby Groups 167
represent two PACs that do not share either a common factor of production (one is
corporate and one labor) or a common industry being represented. The correlation
between the contribution patterns across political candidates for such pairs of
groups is 0.0402. This correlation is significantly larger than zero at the 1%
significance level, with a t-statistic over 62. Thus, even among interest groups
with no clear connection, there is a tendency to give money to the same set of
candidates for the Senate and House of Representatives.
There are several important implications of this result. First, it suggests that
PACs are rational. Snyder (1992) and Stratmann (1992) have provided some
evidence that PAC contributions are motivated by rational considerations of self-
interest, and the results here support their conclusions. If an interest group gave
money to candidates purely because of a utility gain from participating in the
political system, there would be no correlation between the contribution patterns
of different interest groups (particularly between groups that do not represent
either the same factor of production or the same industry). The null hypothesis of
no correlation is emphatically rejected by the data, however, as the model in
Section 7.2 suggests it should be when PACs are motivated by self-interested
factors.
A second implication of the positive correlation between the contribution
patterns of unrelated interest groups is that the electoral motive for giving money
does not dominate the service and influence motives. Suppose for the moment,
that interest groups cared only about which candidates were elected to Congress
and they never sought influence or used contributions as a means of gaining
Table 7.1: Correlations between PAC contribution patterns, by type of PAC.
Standard errors are in parentheses. Numbers of observations in each cell are in brackets.
**Means are significantly different at the 1% level.
C. Magee168
services from politicians. Suppose furthermore that interest groups and candidates
for office were split along factor lines so that each candidate and each PAC could
be considered labor friendly or business oriented. With an electoral motive for
giving contributions, special interest groups with an interest in helping labor-
friendly candidates win election would concentrate their money on labor-friendly
candidates who are engaged in a tight election. Interest groups representing
capital, meanwhile, would focus their contributions on a separate group of
candidates. Thus, we would observe a negative correlation between the
contribution pattern of a labor and a capital interest group, on average.
With an influence motive for giving campaign contributions, on the other hand,
both capital and labor groups would give money to the same set of moderate
candidates who are likely to win election and whose policy positions can be
swayed the most by lobbies with access to the candidate. The service motive gives
all interest groups an incentive to give money to the same group of politically
powerful legislators who are in positions that make them capable of providing
direct aid to lobby groups in the form of setting the legislative agenda, marking up
bills in committees, and pressurizing fellow congress members to vote in a certain
way. The influence and service motivations for campaign contributions, contrary
to the electoral motive, then, predict that unrelated interest groups will tend to give
money to the same set of candidates for office. This prediction is found to have
empirical support.
Table 7.1 reveals that when two PACs represent a common industry, there is a
significant increase in the correlation between their contribution patterns (the
correlation increases to 0.0618). Similarly, if the two interest groups represent a
common factor of production (both are capital or labor), the correlation between
their contribution patterns increases to 0.0667. Finally, if the two PACs share a
common factor and represent the same industry, there is a still greater correlation
between their contribution patterns across candidates (0.1058). These higher
correlations between contribution patterns of PACs when they share more
common characteristics are consistent with the predictions of the theoretical
model in Section 7.2. The model there shows that the more two interest groups
have in common, the greater their incentive to contribute to the same set of
representatives.
Table 7.2 presents the correlations between contribution patterns within PACs
representing capital and those representing labor. There were 20,445 pairs of
interest groups in which one PAC represented capital and the other represented
labor. The average correlation between the contribution patterns of these pairs of
PACs was 0.0424. This mean is significantly greater than zero (the t-statistic is
over 67). When both interest groups were labor unions, the average correlation
between their contribution patterns was 0.0646. This mean was significantly
Industry and Factor Linkages Between Lobby Groups 169
higher ðt ¼ 7:6Þ than the average correlation between contributions from a labor
group and a corporate group. The correlation between two corporate groups’
contribution patterns was 0.0714, which was also significantly higher ðt ¼ 35:4Þthan the average correlation between contributions from a labor group and a
corporate group. Interestingly, two separate corporate lobbies have more similar
contribution patterns on average than two labor unions—the t-statistic examining
the difference in the mean correlations between two corporate lobbies and two
labor lobbies is t ¼ 2:0; indicating that the means are different at the 5%
significance level. It is difficult to judge whether this difference is economically
significant as well. One explanation for this difference is that labor groups seem to
be more likely to give money to challengers than corporate PACs do. Whereas
businesses focus their lobbying efforts on Congress members who have been in
office for a number of terms and have accumulated power in the government, labor
groups do not place such weight on a candidate being a well-entrenched member
of Congress. Magee (2002), for instance, shows that being a committee chair and
an incumbent significantly increased the contributions a candidate received from
business groups in 1996 races for the House of Representatives, but the two
variables had no effect on money received from labor groups.
Table 7.3 examines whether or not corporate and labor PACs have similar
correlation patterns for those groups within the same industry and for those from
separate industries. The first column in Table 7.3 includes only every pair of two
capital PACs, and these pairs of PACs are broken into those that represent
a common industry and those that represent separate industries. The second
column includes only pairs in which both PACs represent the interests of labor.
Table 7.2: Correlations between PAC contribution patterns, by factor types.
Standard errors are in parentheses. Numbers of observations in each cell are in brackets.
*Means are significantly different at the 5% level. **Means are significantly different at
the 1% level.
C. Magee170
The estimates show that corporate PACs that are in the same industry have a
significantly higher correlation between their contribution patterns than corporate
groups representing different industries. For labor groups, however, the average
correlation in contribution patterns between PACs within the same industry is not
significantly different from the average correlation in contributions between PACs
representing separate industries.
Why might there be a greater correlation between PACs within the same
industry for capital groups but not for labor groups? One possible explanation
might be that capital is less mobile across industries than labor, as is commonly
assumed in specific factors models. If labor is perfectly mobile, then industry is
irrelevant in unions’ lobbying efforts. In the Heckscher–Ohlin model with perfect
factor mobility, for instance, industry affiliation is irrelevant for interest groups’
position on trade policy. All groups representing a factor of production either
benefit from free trade and support it politically or are harmed by free trade and
oppose it. Thus, with perfect factor mobility, the relationship between the
lobbying patterns of two interest groups is independent of industry. This
prediction is consistent with the result that the correlation of contribution patterns
between two labor unions within the same industry is no different than the
correlation between two unions from separate industries.
If factors are immobile across industry lines, however, their fortunes rise or fall
with those of their industry. Thus, pressure groups within a particular industry
have a common interest in promoting policies that benefit the industry. The
assumption of immobile factors, then, generates the prediction that groups
representing the same industry will tend to contribute to the same set of policy
makers more often than interest groups representing different industries. This
prediction fits well with the observation that for capital groups there is a much
Table 7.3: Correlations between PAC contribution patterns, by industry type.
Standard errors are in parentheses. Numbers of observations in each cell are in
brackets. **Means are significantly different at the 1% level.
Industry and Factor Linkages Between Lobby Groups 171
higher correlation ðt ¼ 32:3Þ between corporate PACs within the same industry
than between corporate PACs from separate industries.
The implication that capital is less mobile across industries than labor is very
tentative. Examining the correlations between interest group contribution patterns
is not a direct test of mobility, and there are other possible explanations for the
result that corporate groups within industries have more similar contribution
patterns than corporate PACs from separate industries. One possible explanation
is that corporate groups and labor groups have different motivations in giving
campaign contributions. Suppose that firms care primarily about securing services
from politicians on committees relevant to their industry. In this case, defense
industry firms will concentrate their donations on powerful members of national
security committees in the House and Senate while the oil and gas industry
focuses on members of the energy and commerce committee. Thus, there is a
strong correlation between corporate contributions from groups within the same
industry. If labor groups give money primarily with a motivation of helping
certain candidates win election, on the other hand, it is not clear that there should
be a larger correlation in the contribution patterns of labor PACs within the same
industry than there is in the contributions of labor PACs from different industries.
Table 7.4 presents the estimates from a regression in which the dependent
variable is the correlation between two interest groups’ contribution patterns. In
the first column, the determinants of the correlation are assumed to be the factors
examined in Table 7.1: irrespective of whether the two interest groups represented
the same factor and the same industry. In column 2, the factor variable is separated
into two different variables. One is a dummy variable indicating whether or not
both interest groups represent corporate interests (capital ¼ 1, if they do), and the
second is whether or not both interest groups represent labor unions (labor ¼ 1, if
they do). The final column breaks the same industry variable down into separate
dummy variables for each industry.
The results in Table 7.4 are similar to those in Tables 7.1–7.3. Representing the
same factor of production and a common industry, both increase the correlation
between two interest groups’ contribution patterns. The larger coefficient estimate
suggests that representing a common industry is slightly more important than
representing the same factors of production in determining how similar the
contribution patterns of two PACs are. The difference between the two coefficient
estimates in column 1 is statistically significant at the 1% level. Consistent with
the result in Table 7.2, the second column of Table 7.4 shows that two corporate
PACs have more similar contribution patterns than two labor groups do (after
controlling for whether or not the two PACs are in the same industry). When both
interest groups are corporate, the correlation between their contribution patterns
C. Magee172
rises by 0.0285, compared to a 0.022 increase when both groups represent the
interests of labor. These coefficients are significantly different at the 5% level.
In column 3, 15 of the 20 coefficient estimates on the industry dummy variables
are positive, with 11 of these coefficients being statistically significant at the 1 or
5% level. There was a significant negative effect of having two PACs within the
same industry on the correlation between their contribution patterns for only
one industry, food and kindred products. This result is likely to be a Type I error.
The coefficient estimate is very small in magnitude, but because of the large
number of PACs representing this industry, it is statistically significantly different
from zero. Several of the coefficient estimates are quite large. When both PACs
Table 7.4: Determinants of PAC contribution correlations.
Variables Coefficients Coefficients Coefficients
Same factor 0.0284**
Both capital 0.0285** 0.0285**
Both labor 0.0220** 0.0224**
Same industry 0.0363** 0.0363**
Oil and gas 0.0455**
Food 20.0086**
Tobacco 0.3285**
Textiles 0.0501**
Apparel 20.0038
Lumber 0.0643**
Furniture 20.0191
Paper 0.1477**
Printing 0.0295
Chemicals 0.0389**
Rubber 20.0042
Leather 0.2605*
Stone, clay, glass 0.0671**
Primary metals 0.0119*
Fabricated metals 0.0031
Machinery 0.0039
Electronic equipment 0.0415**
Transportation 0.0990**
Instruments 20.0062
Miscellaneous 0.0013
Constant 0.0386** 0.0386** 0.0384**
F-statistic for capital, labor 607.27 615.59
F-statistic for industry variables 157.82
R2 0.0369
Observations 115,921 115,921 115,921
* Coefficients are significant at the 5% level. **Coefficients are significant at the 1% level.
Industry and Factor Linkages Between Lobby Groups 173
are in the tobacco industry, the correlation between their contribution patterns rose
by 0.33. This large estimate may indicate a high degree of coordination between
lobby groups within the tobacco industry, or it may have occurred by random
chance since there were only 11 PACs representing the interests of tobacco in the
data set. There are similarly small numbers of PACs representing the other
industries with very large coefficient estimates, such as the paper (seven PACs)
and leather (two PACs) industries. The industries with the largest numbers of
PACs were the food and kindred products industry, with 128 interest groups
representing it, and the chemicals and allied products, with 105 PACs.
7.4. CONCLUSION
This chapter has presented a theoretical model showing that all interest groups
have incentives to concentrate their campaign contributions on a similar set of
candidates for office. PACs with common policy preferences have even greater
incentives to give money to common groups of candidates. The empirical
evidence supports these predictions. Even among interest groups that did not share
a common factor of production and did not represent the same industry, the
correlation between their contribution patterns in 1991–1992 was 0.04, which was
significantly greater than zero. When PACs represented the same factor of
production, the same industry, or both, the correlation between their contribution
patterns was significantly higher. Interestingly, the contribution patterns of two
corporate lobby groups were more similar, on average, than the contribution
patterns of labor unions.
When factors are partially mobile across sectors, the short-run interests of lobby
groups will depend on the industry they represent while the long-run interests
depend on the factor of production. If trade liberalization harms the textile
industry but raises the overall return to capital in the economy, for example, then
capital owners in the textile industry may suffer short-run losses as the textile
industry declines, but in the long run, after the capital has moved to more
prosperous industries the capital owners will gain from the liberalization.
Lobbying for short-run gains, then, suggests that groups from the same industry
will have the most in common in their contribution patterns. If groups lobby
primarily for long-run gains, then groups representing the same factor of
production will have similar contribution patterns. The results in this chapter
indicate that both short and long-run considerations play a role in determining
interest group contribution decisions, because both representing a common
industry and representing the same factor of production increase the similarity
between two lobbies’ contribution patterns.
C. Magee174
REFERENCES
Beaulieu, E. and Magee, C. (2004). Campaign contributions and trade policy: new tests of
Stolper–Samuelson. Economics and Politics, in preparation.
Grossman, G. and Helpman, E. (1994). Protection for sale. American Economic Review, 84,
833–850.
Grossman, G. and Helpman, E. (1996). Electoral competition and special interest politics.
Review of Economic Studies, 63, 265–286.
Hall, R. and Wayman, F. (1990). Buying time: moneyed interests and the mobilization of
bias in congressional committees. American Political Science Review, 84(3), 797–820.
Hartigan, J. and Tower, E. (1982). Trade policy and the american income distribution.
Review of Economics and Statistics, 64, 261–270.
Magee, S. (1980). “Three simple tests of the Stolper–Samuelson theorem,” in Issues in
International Economics, P. Oppenheimer, (ed.), London: Oriel Press, pp. 138–153.
Magee, C. (2002). Do political action committees give money to candidates for electoral or
influence motives? Public Choice, 112, 373–399.
Magee, S., Brock, W. and Young, L. (1989). Black Hole Tariffs and Endogenous Policy
Theory, Cambridge: Cambridge University Press.
Olson, M. (1965). The Logic of Collective Action, Cambridge: Harvard University Press.
Snyder, J. (1992). Long-term investing in politicians; or, give early, give often. Journal of
Law and Economics, 35, 15–43.
Stratmann, T. (1992). Are contributors rational? Untangling strategies of political action
committees. Journal of Political Economy, 100, 647–664.
Industry and Factor Linkages Between Lobby Groups 175
CHAPTER 8
Sweetening the Pot: How American
Sugar Buys Protection
OMER GOKCEKUSa,*, JUSTIN KNOWLESb, andEDWARD TOWERc
aJohn C. Whitehead School of Diplomacy and International Relations, Seton Hall University,
400 South Orange Avenue, South Orange, NJ 07079, USAbWharton School, University of Pennsylvania, Jon M. Huntsman Hall, 3730 Walnut Street,
Philadelphia, PA 19104-6340, USAcDepartment of Economics, Duke University, Box 90097, Durham, NC 27708-0097, USA
Abstract
We use Probit and Tobit analysis to explore the determinants of campaign
contributions from the sugar industry to Senators from 1989 to 2002. We find that
the power and willingness of Senators to protect sugar influence campaign
contributions significantly: serving on the Senate Agriculture Committee attracts
$4266 of contributions per election cycle; serving on the relevant subcommittee
that deals with sugar legislation is worth an additional $2179 for a total of $6445;
membership of the majority party $1235; and an impressionable freshman Senator
from a sugar cane state receives $7367 more than a more senior senator from a
non-sugar state.
Keywords: Sugar industry, lobbying, trade protectionism
JEL classification: F13
*Corresponding author.
E-mail address: [email protected]
It could probably be shown by facts and figures that there is no
distinctively American criminal class except Congress (Mark Twain,
Pudd’nhead Wilson’s New Calendar).
8.1. INTRODUCTION
Kraft Foods is moving Life Savers production and its 600 jobs, to
Canada. One reason: Sugar is cheaper there. Unlike the US government,
Canada doesn’t prop up prices to protect a handful of domestic sugar
growers (USA Today, 2002).
When economics collides with politics in the halls of Congress, politics
usually wins (Smith, 2002).
The US sugar program is a case in point to show that indeed trade policies
redistribute domestic wealth. In particular it shows what happens when the
beneficiary of a protectionist policy is a small and concentrated group and the
losers are large and widely disbursed all over the country. Sugar growers have
been capturing rents from tariffs and quotas since the 1790s. Taussig (1931)
described sugar protection in his Tariff History of the United States. The US
General Accounting Office (2002) estimated that the sugar program cost
consumers about $1.5 billion in 1996 and about $1.9 billion in 1998. As
Groombridge (2001, p. 1) has written: “nowhere is there a larger gap between the
US government’s free trade rhetoric and its protectionist practices than in the
sugar program.”
A number of researchers have examined the US sugar program. See, for
instance, Harper and Aldrich (1991), who examine the determinants of
congressional voting on the US sugar program, Borrell and Pearce (1999) who
examine the effects of worldwide sugar protection, Lopez (2001), and the articles
in Marks and Maskus (1993).
In this paper, after briefly describing the US sugar program and reporting on
its welfare consequences, we focus on an overlooked aspect of it: the institution
of rent seeking, in particular how the sugar industry—an interest group—picks
legislators to support with campaign contributions. We examine sugar industry
contributions to the reelection campaigns of US Senators for the 14 years from
1989 to 2002. During this period of time, as part of three consecutive Farm
Bills, those in 1991, 1996 and 2001, both Houses voted against phasing out
price supports for sugar, and they extended the federal sugar program. We
attempt to shed light on the relationship between sugar industry contributions
and incumbent Senators’ attributes. In particular, we analyze the premium
O. Gokcekus, J. Knowles, and E. Tower178
attached to the power and willingness of Senators to maintain protectionist
sugar policies, and consequently to supply the rent collected by sugar producers.
Our analysis shows that there is a systematic targeting by the sugar industry in
the allocation of contributions among incumbent Senators. Membership of the
majority party, serving on the Senate Agriculture, Nutrition and Forestry
Committee, and serving on the Senate Agricultural Production, Marketing, and
Stabilization of Prices Subcommittee (which oversees the sugar program) all
attract money.
8.2. THE US SUGAR PROGRAM
The Coalition for Sugar Reform (2003) provides a succinct description of the US
sugar program:
The Federal Government has operated the current price support program
since 1981 to subsidize sugar beet and sugarcane producers and
processors by maintaining high sugar prices. The government supports
the price by restricting the supply of sugar made available to consumers.
It does so by limiting imports. There are no restrictions on domestic sugar
production or marketing. [There are two main components of the
program: Price support loans and import restrictions.]
8.2.1. Price support loans
Under the program, the Government makes loans available to sugar beet and
sugarcane processors. Sugar loans are unique in that they are made to
processors (corporations or cooperatives) rather than to individual farmers, as
under other farm programs. In order to be eligible for a loan, the processor must
pay the producer a Government-specified minimum price for sugar beets or
sugarcane. Processors pledge the sugar as collateral to obtain a so-called “non-
recourse” loan from the Government… When the loan matures, the processor
must decide whether he will make more money by (1) paying off the loan, plus
interest, and redeeming the pledged sugar; or (2) forfeiting the sugar and
keeping the Government’s money… (T)o avoid loan forfeitures… (the US
Department of Agriculture) restricts imports to maintain the…price of sugar
high enough so that processors will have an incentive to redeem every pound of
sugar placed on loan.
Sweetening the Pot: How American Sugar Buys Protection 179
8.2.2. Import restrictions
As mentioned earlier, the Government supports the price of sugar by restricting
imports. It is able to do so because sugar is a deficit crop—we consume more than
we produce. Every quota year (October/September) USDA establishes an overall
Tariff Rate Quota (TRQ) for sugar, which is prorated among some forty nations by
the US Trade Representative on the basis of import history during the period
1975/81. If during the course of the year more sugar is needed, the quota can be
increased. If a nation cannot fill its quota, a deficit can be declared and reassigned
to other quota holding nations… Sugar can enter in excess of the TRQ, but the
importer would have to pay a duty of around 16 cents a pound, which normally
would make it unprofitable.
8.3. CONSEQUENCES OF THE PROGRAM
The Agriculture Committee is writing a new farm bill, and we cannot
afford to have the sugar lobby write the sugar policy. Until the Sugar
Subsidy Program is phased out, costumers will pay more for products
containing sugar. Taxpayers will continue to pay more to buy surplus
sugar.Workers in the candy and the cane refining industry will continue to
lose their jobs. The sugar program will continue to benefit a few, without
solving the problems of family farmers. We must insist on real reform in
the sugar program, and end the regulations that are costing Americans
money and American jobs (Congressman W.O. Lipinski, 2001).
The gap is wide between US and world sugar prices. As Figure 8.1 (drawn from
Table 1 of Appendix I of USGAO, 2002) shows, in 1998 the US raw sugar price
was more than double the world price. Between 1985 and 1998, on average the US
raw sugar price was 3.2 times the world raw sugar price.
Welfare Implications. The US General Accounting Office (2002) recently
examined the US sugar program. We quote their statement of their major findings.
8.3.1. Increases users’ costs
We estimate that the sugar program cost domestic sweetener users about $1.5
billion in 1996 and about $1.9 billion in 1998. Sweetener users included (1)
sugarcane refiners that bought raw cane sugar, (2) food manufacturers that bought
refined sugar and other sweeteners, and (3) final consumers who bought
sweeteners and sweetener-containing products…
O. Gokcekus, J. Knowles, and E. Tower180
8.3.2. Benefits for producers
The primary beneficiaries of the sugar program’s higher prices are domestic sugar
beet and sugarcane producers who, we estimate, received benefits of about $800
million in 1996 and about $1 billion in 1998. About 70 percent of the benefits went
to sugar beet growers and processors. Sugarcane producers received about 30
percent of the benefits.
High Fructose Corn Sweetener (HFCS) producers received little, if any, benefit
from the sugar program in either 1996 or 1998, according to our current model’s
estimates. This result contrasts with our finding in 1993. At that time, HFCS cost a
few cents per pound less than domestic sugar, and both products cost about twice
as much as sugar on the world market… [T]he possibilities for substitution
between sugar and HFCS are more limited than in prior years because
technological advances have improved HFCS products and created more
specialized sweetener markets. As a result, even if the sugar program were
removed and the price of domestic sugar fell substantially, the impact on the price
of HFCS would be limited… Executives from the Corn Refiners’ Association,
which represents HFCS manufacturers, agreed with our model’s results as they
pertained to HFCS, stating that HFCS producers do not benefit from the sugar
program because domestic HFCS prices are no longer linked to sugar prices.
Figure 8.1: The world and US sugar prices (source: GAO/RCED-00-126, p. 14).
Sweetening the Pot: How American Sugar Buys Protection 181
8.3.3. Net effect
We estimate that the sugar program resulted in net losses to the US economy of
about $700 million in 1996 and about $900 million in 1998. Our net loss estimates
include economic inefficiencies and transfers to foreign producers. Economic
inefficiencies occurred, for example, when the sugar program’s artificially high
domestic prices encouraged farmers to grow sugar beets instead of another crop,
such as wheat, that, without the sugar program, might have been relatively more
profitable. Inefficiencies also occurred when artificially high sugar prices
discouraged consumers from purchasing sugar. The cost of these inefficiencies
totaled about $300 million in 1996 and about $500 million in 1998. Transfers from
the US economy to foreign producers occurred because foreign producers received
artificially high prices for the raw sugar they exported to the United States. We
estimate that these transfers amounted to about $400million in both 1996 and 1998.
8.4. SUGAR INTEREST GROUPS’ CONTRIBUTIONS
The industry makes so much money through federal price supports that it
can afford to spend a lot to win political influence in Washington… Just
plain and simple, it’s money to a small group of growers. When you are
getting that kind of money, you can work hard to preserve it. What they
are doing is protecting money, protecting wealth, and protecting higher
incomes (A Capitol Hill Sugar Expert, 1998).
Despite its relatively small size, the US sugar growing industry (henceforth sugar)
is one of Congress’ main campaign contributors. On average, from 1989 to 2002,
sugar contributed $3,090,710 in each election cycle, i.e., $1,545,355 annually, to
incumbents in their campaigns.1Table 8.1 summarizes the total sugar contributions
1 The source of the data is the web page of the Center for Responsive Politics, www.opensecrets.com,
which utilizes reports filed to the Federal Election Committee. They are based on contributions of $200
or more from Sugar grower PACs and individuals to federal candidates and from individual and soft
money donors to political parties, as reported to the Federal Election Commission. Consistently, more
than 90% of the contributions were made by a small number of organizations. For instance in the 107th
Congress, these organizations are Flo-Sun Inc., American Crystal Sugar, American Sugar Cane
League, US Sugar Corp., Southern Minn Beet Sugar Co-op., American Sugarbeet Growers Assn,
Florida Sugar Cane League, Minn-Dak Farmers Co-op., Great Lakes Sugar Beet Growers, Florida
Sugar Cane League, Snake River Sugar, Sugar Cane Growers Co-op. of Florida, US Beet Sugar Assn,
Rio Grande Valley Sugar Growers, and Amalgamated Sugar. The contributions are in real terms, i.e., in
107th congress dollars. We use the consumer price index (CPI relative to the average CPI in 2001 and
2002) as the deflator.
O. Gokcekus, J. Knowles, and E. Tower182
for the last seven congresses, i.e., The 101st–107th Congresses. In addition to the
total contributions, this table gives the type of contributions, i.e., individual, PACs,
and Soft Money. It also shows how the total contributions were allocated between
two parties. An examination of the aggregate sugar campaign contributions since
1989 reveals interesting patterns.
First, clearly, soft money has become more important. This is an interesting
pattern because “soft money” contributions are not subject to the limits and
regulations of federal election laws. Soft money contributions are described as the
main vehicle for wealthy individuals, corporations, and labor unions to deliver
millions of dollars to political parties without regard for the limits and other
requirements of the law (Center for Responsive Politics, 1998). Sugar has utilized
this vehicle intensively. Second, incumbent representatives from the Democratic
Party have received 57% of the contributions to incumbents.
However, as is presented in Figure 8.2, whether the party is majority or not also
played a role. Our regression analysis shows that, when not in the majority, the
Democratic Party attracted 50% of contributions and the Republican Party
attracted 38%, with 12% more attracted by the majority party, which is either the
Democratic or Republican party. This means that sugar does take into account the
potential influence ( power) of a party in allocating its contributions.
There is a substantial literature in the political science and public choice areas
on campaign contributions and how interest groups allocate the contributions
among legislators, (e.g., Munger, 1989; Grier et al., 1990). In Section 8.5, we
adopt a model, which is built on the premise that the size of contributions for
reelection depends on a legislator’s characteristics. In particular, we explore the
relationship between campaign contributions and the incumbent Senators’ party
affiliation, state, seniority, relevant committee and subcommittee memberships.
Table 8.1: Campaign contributions by sugar interest: 1989–2002.
Congress Total (in $) Individual (%) PACs (%) SOFT (%) To Democrats (%)
101st 2,571,056 9 91 NA 60
102nd 2,729,790 17 71 11 67
103rd 2,841,146 9 79 12 65
104th 3,916,350 12 61 26 46
105th 3,059,715 9 65 25 52
106th 3,561,589 9 47 44 52
107th 2,955,323 8 67 25 56
Totals are in real terms, i.e., in 107th congress dollars. The deflator is the CPI. All tables, figures, and
discussion refer to a typical two-year election cycle.
Sweetening the Pot: How American Sugar Buys Protection 183
8.5. DETERMINANTS OF SUGAR’S CAMPAIGN CONTRIBUTIONS
In this section we check the association between campaign contributions and
different characteristics of an incumbent Senator. We are effectively testing the
validity of four prior assertions, which were developed by reviewing the literature
on campaign contributions by interest groups to incumbent Senators.
First, we examine the relationship between campaign contributions and party
affiliation. To capture both Senators’ power and willingness to provide what sugar
interests want, we conduct two sets of analyses. We examine the contributions to
Democratic and Republican Senators. As Table 8.2 shows, incumbent Senators
affiliated with the Democratic Party received 43% more than incumbent
Republican Senators. We initially took this as an indicator of the premium
attached to the “ideology or reputation for particular policy belief”, e.g.,
Democrats are protectionist and Republicans support market solutions. But this
finding contrasts with the results from our more sophisticated analysis presented
below. That analysis uses probit and Tobit regressions to show that correcting for
other factors, party does not matter. It demonstrates the importance of model
building to understand how campaign contributions are allocated.
Figure 8.2: Allocation of sugar campaign contributions between two political parties. Results from
regressing each party’s Share of Contributions over the seven Congresses (14 observations) on
Majority Party, which is one for the ruling party in the Senate and zero otherwise, and Democrat, a
dummy which is 1 or 0 depending on whether the party is Democratic. The intercept is 38%, and both
coefficients of the Democratic and Majority Party dummies are 12%.
O. Gokcekus, J. Knowles, and E. Tower184
We also examine the contributions to incumbent Senators according to their
membership of the Majority party: Incumbent Senators who were members of the
majority party received 40% more than incumbent Senators in the minority party.
We interpret this as an indicator of the premium collected for the “institutional
power or productivity,” of the Senator due to his party affiliation.
Second, we explore the relationship between campaign contributions and the
state a Senator represents. We hypothesize that the Senators who have a large
presence of sugar producers in their states receive larger campaign contributions
than the other Senators, because it is easier to get them to vote on behalf of sugar,
i.e., they are more willing. In particular, we check to see if there is a difference
between Senators from the four sugar cane states—Florida, Hawaii, Louisiana,
Table 8.2: Determinants of campaign contributions by sugar interest to incumbent Senators.
Average(in 2002 dollars)
Percent ofSenators received
Average(among those who received)
Party
Democrat $6061 52% $11,661
Republican $4230 46% $9121
Ratio 1.43 1.12 1.28
Sugar states
Sugar Cane State $8923 71% $12,583
Not a Sugar state $3979 42% $9412
Ratio 2.24 1.68 1.34
Sugar Beet $7643 64% $11,933
Not a Sugar state $3979 42% $9412
Ratio 1.92 1.52 1.27
Majority Party
Majority Party $5940 52% $11,405
Minority Party $4230 46% $9219
Ratio 1.40 1.14 1.24
Committees
Agr. Committee $10,778 69% $15,568
Not Agr. Comm. $3889 45% $8693
Ratio 2.77 1.55 1.79
Sub-committee $11,722 75% $15,629
Not Agr. Comm. $3889 45% $8693
Ratio 3.01 1.68 1.80
Seniority
Freshman $6800 61% $11,068
Sophomore $5683 49% $11,650
Junior $4508 42% $10,805
Senior þ $2687 38% $7023
All members $5168 49% $10,487
Sweetening the Pot: How American Sugar Buys Protection 185
and Texas, and Senators with no sugar industry among their constituency; and
whether there is a difference between Senators from the eleven Sugar Beet
States—California, Colorado, Idaho, Michigan, Minnesota, Montana, Nevada,
North Dakota, Oregon, Washington, and Wyoming, and Senators with no sugar
industry among their constituency. As Table 8.2 shows, Senators from sugar cane
states receive 124% more, and Senators from sugar beet states receive 92% more
than the Senators with no sugar industry within their constituency, with the
difference presumably reflecting the smaller size of the sugar beet sector.
Third, we assess the relationship between campaign contributions and seniority
in the Senate. Seniority could be seen as a proxy for electoral security and also
procedural expertise and collegial respect. There is empirical evidence that
freshmen Senators are perceived overall as being less secure and they usually
receive more than the others (Endersby and Munger, 1992; Grier et al., 1986;
Grier and Munger, 1993). Sugar gives almost 50% more money per capita and
more frequently to freshman Senators than the others.
Finally and fourth, we explore the relationship between committee and also
subcommittee memberships and campaign contributions. “Each committee has
significant power to veto, or at a minimum delay substantially, legislation within
its jurisdiction, to an extent determined by specific rules of the Senate, particularly
when committee control over the conference stage of bills is taken into account.
Committees, therefore, have both agenda and proposal control denied to
nonmembers. Further, even after a bill is passed and is being administered or
regulated by the bureaucracy, the committee with oversight jurisdiction has
substantial power to influence the administrative agency through hearings, agency
appointments, and appropriations.” (Grier and Munger, 1993, p. 619). In
particular, we focus on the Agriculture, Nutrition and Forestry Committee and
the Production and Price Competitiveness Subcommittees.
“The Senate Committee on Agriculture, Nutrition, and Forestry has helped
establish, guide, and examine agricultural policies here and abroad. It has had a
hand in fashioning the research and teaching of the 1860s, the price and income
support controls of the 1930s, and the international trade of the 1990s.” (US
Government Printing Office, 1998, p. 2) In particular, as part of the Farm
Bill, every five years, the committee revisits the sugar price–support program. As
a result of the division of labor within the committee the sugar price–support
program is in the responsibility area of the Production and Price Competitiveness
Subcommittee.2 As Table 8.2 presents, the committee members receive 177%
2 Until the early 1990s, the name of the same subcommittee was Agricultural Production, Marketing,
and Stabilization of Prices.
O. Gokcekus, J. Knowles, and E. Tower186
more; and subcommittee members receive 201% more than non-committee
members.
These preliminary cross-tabulations provide supporting evidence on all four of
our prior assertions. Clearly, sugar knows what it is doing: make the most out of
the money contributed by a systematic targeting according to Senators’ power and
willingness to provide what sugar wants.
To focus on and magnify this pattern, we further divide our sample into groups.
As Figure 8.3 shows, sugar targets the powerful: a non-committee member
minority party Senator receives only $3329 from sugar; a Senator from the
majority party who is also a member of the subcommittee receives 4.06 times as
much, i.e., $13,530 in an election cycle.
Figure 8.4 summarizes another salient future of targeting by the sugar industry,
namely targeting willing incumbent Senators: sugar contributes more to both
incumbent Senators from a sugar cane state and to freshmen Senators. For
instance, a freshmen Senator from a sugar cane state receives $24,009. On the
other hand, an incumbent Senator who is neither a freshman nor from a sugar state
receives only $3567.
The analysis in this section shows that indeed, Senators’ attributes matter—they
impact the contributions from the sugar industry to the reelection campaigns of the
incumbent Senators. However, it does not answer the following critical questions:
Figure 8.3: Sugar contributions and Power.
Sweetening the Pot: How American Sugar Buys Protection 187
How much does each one of these attributes impact campaign contributions?
In other words, what is the marginal impact of different attributes on campaign
contributions?
Accordingly, Section 8.6 represents an attempt to move beyond this section’s
finding that attributes of Senators matter, and to open up consideration of partial
effects of different attributes. In other words, we examine the effect of attributes,
holding other attributes constant. First, we examine the partial impact of Senators’
attributes on the probability of receiving contributions from the sugar industry.
Second, we examine the partial impact of Senators’ attributes on the amount of
contributions received from the sugar industry.
8.6. MARGINAL EFFECTS OF DIFFERENT ATTRIBUTES ON THE
PROBABILITY OF GETTING MONEY AND THE AMOUNT OF
MONEY RECEIVED BY INCUMBENT SENATORS
First, we ask the simple question: How do different attributes affect the probability of
getting contributions fromsugar?Toanswer this questionwe estimate a probitmodel.
An incumbent Senator either receives contributions ðC ¼ 1Þ or does not ðC ¼
0Þ in an election cycle.3As is argued in Section 8.5, we believe that each Senator is
Figure 8.4: Sugar contributions and Willingness.
3 For details, see Greene, 2003, pp. 665–668.
O. Gokcekus, J. Knowles, and E. Tower188
unique and possesses a certain combination of attributes. We believe that a set of
attributes, such as PARTY, FRESHMAN, SUGAR CANE STATE, SUGAR
BEET STATE, COMMITTEE, and SUBCOMMITTEE, gathered in a vector x
explain the decision of the sugar industry whether to contribute, so that
ProbðC ¼ 1Þ ¼Fðx;bÞ;
ProbðC ¼ 0Þ ¼12 Fðx;bÞ;
where the set of parameters b reflects the impact of x on the probability.4
Briefly, REPUBLICAN is dummy variable, which takes on the value of 1 for a
Republican and the value of 0 for others. The variable, MAJORITY PARTY, is
also a dummy variable, which takes the value of 1 for a Senator who is a member
of the majority party during the election cycle and the value of 0 for others.5
FRESHMAN dummy takes on the value of 1 for a Senator who is in her first term.
SUGAR CANE STATE dummy takes the value of 1 for Senators who are from
four states with significant sugar cane production. These states are Florida,
Hawaii, Louisiana, and Texas. Similarly, SUGAR BEET STATE dummy variable
takes the value of 1 for Senators who are representing eleven states with a
significant sugar beet production. These states are California, Colorado, Idaho,
Michigan, Minnesota, Montana, Nevada, North Dakota, Oregon, Washington, and
Wyoming. Finally, the vector x includes two Committee membership dummy
variables: AGR. COMMITTEE dummy takes the value of 1 for Senators who are
members of the Senate Agriculture, Nutrition and Forestry Committee, and 0
otherwise; and PR. SUBCOMMITTEE dummy takes the value of 1 for Senators
who are members of the Senate Agriculture, Nutrition and Forestry Committee
and serve on the Agricultural Production, Marketing, and Stabilization of Prices
Subcommittee.
Table 8.3 presents the estimated coefficients and marginal effects for the probit
model.6 The model predicts that changing party affiliation from Republican to
Democratic increases the probability of getting contributions from sugar by 5%
points. Being a member of the majority party increases the probability of getting
contributions by 4% points. A freshman has a 17% point higher probability of
4 We use the normal distribution in our analysis, i.e., ProbðC ¼ 1Þ ¼Ðb0x21
fðtÞdt ¼ Fðb0xÞ ¼ Fðx;bÞ;
where F(·) is the notation for the standard normal distribution.5 Democrats were the majority for four of the congress, namely the 101st, 102nd, 103rd, and 107th; and
Republicans were the majority three times: the 104th, 105th, and 106th Congresses.6 To capture the potential impact of the number of years in Senate, we included an incumbency variable
in various functional forms, e.g., linear or second order polynomial. We could not get any significant
results. Therefore, for simplicity, we decided not to include years of incumbency in the model that we
present.
Sweetening the Pot: How American Sugar Buys Protection 189
getting money for her reelection campaign from sugar. A Senator from a sugar
beet state has a 20% point, and a Senator from a sugar cane state has a 33% point
higher probability of getting contributions.
Finally, in their reelection campaigns members of the Senate Agriculture,
Nutrition and Forestry Committee have a 17% point higher probability of getting
money. If a member of this committee is also a member of the Agricultural
Production, Marketing, and Stabilization of Prices Subcommittee, the probability
of getting sugar money goes up by another 19% points.7
Although the findings of these estimation results are revealing, they do not
tell the whole story: first, the contributions show variations within a wide
range. Although the mean and median are $10,516 and $7229, respectively,
the maximum and the minimum amounts of contributions are $68,793 and
$129, respectively. (The standard deviation is $3441.)
Those who do not receive money cannot be put into the same category as
those who do: even if the sugar interest hates an incumbent Senator, and it
Table 8.3: Probit model maximum likelihood estimates: probability of receiving sugar campaign
contributions in each two-year election cycle.
Variable Coefficient ðbÞ t ratio Marginal effect t ratio
Constant 20.463 24.43
Republican 20.128 21.28 20.050 21.32
Majority Party 0.104 1.04 0.041 1.03
Freshman 0.442 4.13 0.174 4.14*
Sugar Beet State 0.499 4.07 0.196 4.05*
Sugar Cane State 0.848 4.37 0.334 4.31*
Agr. Committee 0.427 2.30 0.168 2.30*
Pr. Subcommittee 0.494 2.08 0.194 2.08*
Frequencies of actual and predicted outcomesa
Predicted
Received ¼ No Received ¼ Yes Total (actual)
Received ¼ No 257 98 355
Actual Received ¼ Yes 140 205 345
Total (predicted) 397 303 700
Notes: Log L ¼ 2440:27; x2 ¼ 89:72: Marginal effects are computed at the means of the explanatory
variables. *Based on a one-sided t-test, statistically significant at 1% level.aThe model predicts 66% of the recipients accurately (462 out of 700). In particular, the model
accurately predicts 257 out of 355 of those who did not receive; and 205 out of 345 of those who
received.
7 We should keep in mind that these marginal effects are only suggestive because, marginal effects are
computed at the means of the explanatory variables. Endnote 10s caveats also apply.
O. Gokcekus, J. Knowles, and E. Tower190
would like to punish him by taking money out of his campaign (or
contributing negative dollars to his reelection campaign), this is not possible.
To deal with this issue, while estimating the impact of different attributes of
incumbent Senators on the amount of money they receive from sugar, we
propose the following Tobit model:
We follow convention and assume that there is a latent variable Cp; desiredcontribution, which could be negative, and it is linearly related to the attributes of
Senators, and that Cp; is observed only when Cp; is positive (because actual
contributions must be non-negative).8 In other words, if the desired contribution is
negative, we only observe a zero contribution. Accordingly,
Cp¼ b0
xþ e ;
and
C ¼Cp; if b0
xþ e . 0;
0; otherwise:
(
Table 8.4 presents the estimated coefficients and marginal effects for the Tobit
model. The model predicts that a Republican Senator gets only $9 less than a
Democratic Senator, per two-year election cycle. In other words, party affiliation
does not make an important difference.9 However, if a Senator is a member of the
majority party, the sugar industry contributes $1235 more than for a minority
party member. An impressionable freshman Senator gets significantly more: a
freshman Senator receives an additional $2181 from sugar. A Senator from a sugar
beet state receives an additional $2982, and a Senator from a sugar cane state gets
an extra $5187 in campaign contributions.
8 For details, see Greene, 2003, pp. 762–766.9 Each b coefficient in Table 4 shows the effect on Cp of a change in an x variable. For example,
switching from being a Democrat to a Republican decreases desired contributions, Cp; by $19. Each
marginal effect shows the effect on C of an incremental change in an x variable, for an individual whose
initial xs are the sample means. In other words, each marginal effect reckons with the probability of
receiving a contribution being less than one. Thus, loosely speaking, the marginal effects show the
effect of a change in each x variable for a “typical” Senator. However, since all x variables are
dummies, there is no “typical” Senator: for example, no one is half a Republican and half a Democrat.
Moreover, since dummies take the values of only one or zero, in the real world there is no such thing as
an incremental change in one of the xs: Bearing these caveats in mind, the reader should glean from
Table 8.4 that if the “typical” Senator switches from being a Democrat to being a Republican, he loses
$9 in campaign contributions.
Sweetening the Pot: How American Sugar Buys Protection 191
Finally, per two-year election cycle, a member of the Senate Agriculture,
Nutrition and Forestry Committee gets $4266 more than a Senator who is not a
committee member. If a member of this committee is also a member of the
Agricultural Production, Marketing, and Stabilization of Prices Subcommittee
that translates into additional $2179 of sugar money. In other words, serving on
the Subcommittee reaps an additional $6445 contribution (compared to a Senator
who is not a member of the Agriculture Committee).
8.7. CONCLUDING REMARKS
It is revealing to examine the systematic way contributions from the sugar industry
are directed to the reelection campaigns of incumbent Senators. It explains the
longevity of the US sugar program despite its huge welfare losses. It is not a
surprise to see how Sugar growers have been capturing substantial rents from
tariffs and quotas. A close examination of the determinants of campaign
contributions to Senators from the sugar industry from 1989 to 2002 reveals
interesting points. We find that the power and willingness of the Senators to
support sugar influence the campaign contributions significantly: Serving on the
relevant subcommittee that deals with sugar legislation is more profitable than
serving on the agriculture committee alone. Serving on the Senate Agriculture,
Nutrition and Forestry Committee attracts $4266 of sugar contributions per
election cycle, but serving on that committee plus the Agricultural Production,
Marketing, and Stabilization of Prices Subcommittee is worth $6445. These
results suggest the strength of subcommittees in drafting specialized legislation
and attracting interested members. Moreover, Tobit analysis also shows that while
Table 8.4: Tobit model maximum likelihood estimates: sugar campaign contributions to incumbent
Senators in each two year election cycle.
Variable Coefficient ðbÞ t ratio Marginal effect t ratio
Constant 29001.33 27.37
Republican 219.43 21.37 28.95 21.37*
Majority Party 2681.19 2.18 1235.27 2.18**
Freshman 4733.73 3.68 2180.91 3.69***
Sugar Beet State 6472.91 4.45 2982.18 4.46***
Sugar Cane State 11,257.58 5.20 5186.56 5.20***
Agr. Committee 9259.54 4.25 4266.03 4.24***
Pr. Subcommittee 4730.36 1.78 2179.36 1.78**
Log L ¼ 23996:77; N ¼ 700: *Based on a one-sided test t-test, statistically significant at 10% level.
**Based on a one-sided test t-test, statistically significant at 5% level. ***Based on a one-sided test t-
test, statistically significant at 1% level.
O. Gokcekus, J. Knowles, and E. Tower192
the party affiliation does not make any difference, membership of the majority
party is worth $1235. Finally, an impressionable, insecure freshman Senator
receives $2181 more.
So, are campaign contributions productive? As mentioned above, in 1998, the
sugar program transferred roughly $1 billion dollars to sugar growers. From Table
8.1, in the 1997–1998 election cycle they contributed $3,059,715, adjusted for
inflation. We reverse the inflation adjustment and divide the contribution by two
to put it on an annual basis. We discover that sugar producers receive a reward
of 714 dollars for each dollar they spend on electoral campaign contributions!
ACKNOWLEDGEMENTS
Knowles started this project as part of his senior honors thesis and masters’ thesis
at Duke. Holger Sieg and Tower were the advisors. Gokcekus refined and updated
the project, and drafted the present version of the paper. We would like to thank
the following for their help: John Aldrich, Jesus Araiza, Charles Cording, Julia
Cormano, John Brehm, Erin Fletcher, Paul Gronke, Craufurd Goodwin, Paul Izzo,
Bill Kaempfer, Stephan Kretzchmar, Sheila Krumholz, Michael Munger, David
Price and Holger Sieg, Brandon Shapiro, and Nathan Spanheimer.
REFERENCES
ACapital Hill Sugar Expert (1998). Quoted in The Politics of Sugar: White Gold, Center for
Responsive Politics. www.opensecrets.org/oubs/cashingin_sugar/sugar01.html.
Borrell, B. and Pearce, D. (1999). Sugar: The Taste Test of Trade Liberalization. Centre
for International Economics. www.intecon.com.au/pdf/Sugar_taste_test_of_trade_
lib_report.pdf.
Center for Responsive Politics (1998). The Politics of Sugar: White Gold. www.
opensecrets.org/oubs/cashingin_sugar/sugar01.html.
Coalition for Sugar Reform (2003). How the Government’s Sugar Subsidy Program Works.
www.sugar-reform.org/Sugprog.htm.
Endersby, J. W. and Munger, M. C. (1992). The impact of legislator attributes on union
PAC contributions. Journal of Labor Research, 8, 79–97.
Greene, W. H. (2003). Econometric Analysis, 5th edn., Upper Saddle River: Prentice Hall.
Grier, K. andMunger, M. C. (1993). Comparing interest groups PAC contributions to house
and senate incumbents, 1980–1986. The Journal of Politics, 55, 615–643.
Grier, K., Munger, M. C. and Roberts, B. (1986). The determinants of industry political
activity, 1978–1986. American Political Science Review, 88, 911–926.
Sweetening the Pot: How American Sugar Buys Protection 193
Grier, K., Munger, M. C. and Torrent, G. M. (1990). Allocation patterns of PAC monies:
the US senate. Public Choice, 67, 111–128.
Groombridge, M. A. (2001). America’s Bittersweet Sugar Policy. Cato Institute Trade
Briefing Paper No. 13.
Harper, R. K. and Aldrich, J. (1991). Political economy of sugar legislation. Public Choice,
70, 299–314.
Lipinski, W. O. (2001). Special Order Speech on Sugar Program Reform. www.house.gov/
lipinski/wh_sugarreform.htm.
Lopez, R. A. (2001). Campaign contributions and agricultural subsidies. Economics and
Politics, 13, 257–279.
Marks, S. V. and Maskus, K. E. (eds.), (1993). The Economics and Politics of World Sugar
Policies, Ann Arbor: The University of Michigan Press.
Munger, M. (1989). A simple test of the thesis that committee jurisdictions shape corporate
PAC contributions. Public Choice, 62, 181–186.
Smith, G. (2002). Profile: Tax Dollars, Higher Prices at Market Sweeten Deal for Some Big
Sugar Producers at Our Expense, NBC Nightly News—02/14/2002 (an interview).
Taussig, F. W. (1931). The tariff history of the United States, New York: Macmillan
Company.
US General Accounting Office (2002). Sugar Program: Supporting Sugar Prices Has
Increased Users’ Costs While Benefiting Producers, GAO/RCED-00-126. http://faculty.
washington.edu/belas/543/GAOSugar.pdf.
US Government Printing Office (1998). The United States Senate Committee on
Agriculture, Nutrition, and Forestry: 1825–1998: Members, Jurisdication, and History,
S. Doc. 105-24.
USA Today (2002). Sugar Lobby’s Clout Threatens Economic Decay, August 19.
O. Gokcekus, J. Knowles, and E. Tower194
CHAPTER 9
Unilateralism in Trade Policy:
A Survey of Alternative
Political-Economy Approaches
PRAVIN KRISHNA,a DEVASHISH MITRA,b,*
aDepartment of Economics, Brown University, Box B, 64 Waterman Street, Providence, RI 02912, USAbDepartment of Economics, The Maxwell School of Citizenship and Public Affairs, 133 Eggers Hall,
Syracuse University, Syracuse, NY 13244, USA
Abstract
While much of the literature has considered the merits of the two policy stances of
unilateralism and reciprocity independently of each other, we focus on the
possible causal connection between the two. Can unilateral trade liberalization by
one country lead to reciprocal liberalization by its partner in the absence of
negotiations between the two countries? In this survey paper, we consider this
issue analytically in the context of different political economy approaches to trade
policy determination.
Keywords: Trade policy, political economy, lobby formation, unilateralism,
reciprocity
JEL classifications: F10, F13, F02
*Corresponding author.
E-mail address: [email protected]
9.1. INTRODUCTION
The relative merits of two alternative approaches to progressing towards free
trade, unilateralism (I drop my trade barriers) and reciprocity (I drop my trade
barriers if you drop your trade barriers), have long been debated in the theoretical
literature on trade policy.
The arguments used to support these policy stances are well known. The
argument for unilateral trade liberalization has relied on the demonstration that, in
the absence of “distortions”, free trade is welfare maximizing. Further, even in the
presence of a very wide variety of distortions in the system, the merits of a policy
of free trade (coupled usually with non-trade interventions) have been decisively
established.1 A policy stance of reciprocity, on the other hand, relies on the large
country assumption that leads to directly trade-related distortions in the form of
“terms-of-trade” effects with or without political-economy forces in the economy.
Thus, for instance, Mayer (1981) shows that in the presence of terms-of-trade
motivations for tariffs, international negotiations could lead to a better outcome
than the non-cooperative Nash outcome derived earlier by Johnson (1953).
Equally, political economy influences have been considered in models explaining
agreed-upon reciprocal trade liberalization in the work of Mayer (1984a), Hillman
and Moser (1996), and Bagwell and Staiger (1999), among others.2
While most of the literature has considered these two approaches to trade
liberalization independent of each other, we have, in Krishna and Mitra (2003a,b),
studied their possible causal interaction.3 We have found that unilateral trade
liberalization in one country can induce reciprocal liberalization by another
country in the absence of any communication or negotiations between the two
countries. This causal linkage between unilateralism and reciprocity is explored
in these two papers in the context of two different models of trade policy
determination. The first (the one in Krishna and Mitra (2003a)) is a framework in
which trade policy changes are voted upon by voters and where outcomes are
determined by the majority (as, for instance, in Fernandez and Rodrik (1991)).4
In this context, we demonstrate that unilateral trade liberalization by a trading
partner increases the electoral support for trade liberalization in the home country.
1 See Johnson (1965) and the classic generalized treatment of Bhagwati (1971).2 In this context, see also Grossman and Helpman (1995), where they analyze the merits of “trade talks”
over “trade wars” in a two-country, specific-factors setting with lobbying taking place within each
country.3 See Bhagwati (1990) for an early informal discussion of this idea.4 See also Mayer (1984b) for the first median-voter model of trade policy determination, both in
Heckscher–Ohlin and specific-factor contexts.
P. Krishna and D. Mitra198
The second framework (Krishna and Mitra, 2003b) is the popular Grossman
and Helpman (1994) model of trade policy determination in which the policy in
the home country is assumed to be determined through the interaction between the
government and organized lobbies representing economic interests. We augment
this with the lobby formation framework of Mitra (1999). In this context too, we
show that unilateral liberalization by one country can influence another country to
liberalize its trade regime through the formation of an export lobby there.
Coates and Ludema (2001), in an important paper, study the impact of unilateral
tariff reduction on negotiation outcomes (specifically the likelihood of success
in reaching and ratifying bilateral agreements in the presence of the “political risk”
of domestic opposition to trade agreements) and argue that unilateral trade
liberalization may be the optimal policy for a large country. In their framework,
“unilateral liberalization acts as insurance” by providing a “risk-sharing” role.
Unilateral tariff reduction lowers the political stakes associated with trade
liberalization in the foreign country, thereby lowering the overall political cost of
reaching and implementing trade agreements and increasing the probability of
successful agreements. While our own work in this area relies on perfectly non-
cooperative interaction between trading partners, Coates and Ludema use a more
cooperative setting where negotiations take place between countries.
Whatever the political-economy channel, the possibility that unilateral
liberalization induces liberalization by the partner suggests a greater case for a
policy stance of unilateralism than is traditionally recognized in the literature. We
believe that this result carries important normative implications. Specifically, it
challenges the frequently proposed strategy of the use (or the threat of the use) of
raising one’s trade barriers to remove those of others, for example, the United
States’ recent use of the “Super-301” provision to retaliate against “unreasonable”
trade barriers by raising its own.5
The preceding discussion has focused on the possibility of unilateral
liberalizations inducing reciprocity in a partner country. While countries may
unilaterally liberalize with this end in mind, they may also do it for other reasons.
We round out our discussion by briefly considering some political-economy
models of endogenous unilateralism (Maggi and Rodriguez-Clare, 1998; Mitra,
2002). While these papers have not focused on the issue of reciprocity by partners,
5 The policy recommendations that follow from this analysis of unilateralism and reciprocity have
been elegantly summarized by Bhagwati (2002) as follows (1) “Go alone (that is, cut trade
barriers unilaterally) if others will not go with you”, (2) “If others go simultaneously with you
(i.e., there is reciprocity in reducing trade barriers), that is still better”, and finally (3) “If you
must go alone, others may follow suit later: unilateralism then leads to sequential reciprocity”.
Unilateralism in Trade Policy 199
reciprocity of the type we have discussed may follow even in the contexts that
they analyze.
The rest of this chapter proceeds as follows. In Section 9.2 we present first a
non-technical discussion of political-economy mechanisms driving what we call
“reciprocated unilateralism” in trade policy. The first model is a median-voter
model, while the second is a Grossman–Helpman lobbying model augmented
with lobby formation. We end this section with a brief discussion of the Coates–
Ludema model. We then discuss in Section 9.3 some political-economy models
which have demonstrated the possibility of endogenous unilateralism. In Section
9.4 we present some concluding observations.
9.2. MODELS OF RECIPROCATED UNILATERALISM
9.2.1. A median-voter model
We, here, provide a very non-technical and abridged version of the analysis in
Krishna and Mitra (2003a). Consider an economy with two sectors, M (import
competing) and E (exportable). We assume that both goods in this economy are
produced under constant-returns to scale using labor alone. However, different
individuals have different levels of productivities in the production of the two
goods. We assume that prior to the trade reform, there is a positive import tariff in
place, which goes down to zero if and when the trade reform is implemented. We
also assume that a personworks in the sector that pays her a higherwage determined
by her productivity. Thus, every individual in the economy has his or her
comparative advantage in the production of the export good relative to producing
the import-competing good, given by the ratio of the person’s productivity in the
export sector relative to that in the import-competing sector. This measure of
comparative advantage needs to be greater than the relative price of the import-
competing good for an individual to decide to work in the export sector. Thus given
the relative price of the importable, we can determine who andwhat fractions of the
people will work in the two sectors. In equilibrium, the marginal worker will be
indifferent between working in the export and import sectors.
We are unambiguously able to show that individuals who were originally in the
exportable sector will benefit from these reforms. All individuals who were
originally in the M sector and remain there in the post reform equilibrium are
clearly shown to be worse off. Finally, we look at the individuals who were in the
import-competing sector prior to the reforms but are in the export sector in the
post-reform equilibrium. In other words, these are individuals that end up moving
to the relatively more lucrative sector for them after the reforms. Note that given
P. Krishna and D. Mitra200
the post-reform relative prices, the export sector might be relatively more
lucrative but some of these people may be worse off relative to their pre-reform
situation in which the domestic relative price was different. After taking this into
account, we see how the overall support (the number of people who were
originally in the export sector plus the people changing their profession who
benefit from these reforms) for reforms responds to the world relative price of
imports and to the size of the original tariffs. Not surprisingly, the support for the
reform is decreasing in the world relative price of imports as well as the initial
tariff. This is intuitive since for given relative productivity and tariff, the potential
relative wage of an individual in the import sector (relative to that in the export
sector) is increasing in the world relative price of the importable. And so if
majority voting is the mechanism through which decision on the reforms is made,
it takes place when the world price is below a certain threshold price.
For a small country, the world price, though independent of its own tariff, can
depend on and be increasing in the tariff of a large trading partner. Thus, if such a
partner country reduces its tariff, the world price of the importable will go down
(equivalent to the relative price of the exportable going up) and the support for a
trade reform in this partner country will go up. The intuition is that with a lower
world relative price of the importable, the relative wage advantage (disadvantage)
from working in sectorM is lower (higher) for any given initial tariff. This reduces
(increases) the attractiveness of being in the import-competing sector without a
trade reform.
We next consider the case of two large open economies trading with each other.
In this case, the world price will be a function of the tariffs of the two countries, and
therefore, a trade reform in the partner country raises support for reforms in the
home country. If foreign tariff is very high, the support for reform at home is small
and so there is no reform and this country is stuck at the intial tariff as reforms
cannot take place. For a foreign tariff below a critical level, there is majority
support for the reform and the home tariff goes down to zero. Analogously, for a
home tariff below a critical level, there is majority foreign support for a reform in
their country and the foreign tariff goes down to zero. Thus, there is the possibility
of multiple equilibria—either both countries liberalize or both countries remain
stuck at their respective initial tariffs. Therefore, a dictator or an international
organization forcing reform in one country can, indirectly (through majority
voting), bring about a reform in the partner country. This situation can then
become a new equilibrium consistent with majority voting in both countries.
The above analysis is also generalizable to the case with endogenous tariff
levels and to the introduction of individual-specific uncertainty of the type
analyzed in Fernandez and Rodrik (1991). The results described above remain
qualitatively unchanged.
Unilateralism in Trade Policy 201
9.2.2. A model with trade policy lobbying and endogenous lobby formation
The development here follows Krishna and Mitra (2003b) very closely but the
presentation is much more non-technical. Consider a small open home economy,
exactly as in Grossman and Helpman (1994), producing a numeraire good using
labor with Ricardian technology and two non-numeraire goods (an import-
competing and an exportable good), each requiring a different kind of factor of
production specific to that good and labor for their production. Individuals in this
economy are assumed to have identical preferences represented by utility
functions linear in the consumption of the numeraire good, strictly concave in the
consumption of each of the other two goods and additively separable in the
consumption of all three goods.
First, we assume that each individual in the economy is endowed with exactly l
units of labor. Further, we assume that each individual owns only one type of
specific factor and that owners of any particular type of specific factor are
symmetric (that is, they own identical amounts of that specific factor).
For given lobbies, the trade policy vector is exactly the same as in Grossman
and Helpman (1994). When only the import-competing sector is organized, there
is an import tariff in place for this sector and there is an export tax on the other
non-numeraire sector. Given our assumptions above (in particular that the
assumption that the fraction of the population owing either specific factor sums to
one), it is well known that the Grossman–Helpman trade tax vector would equal
the zero vector if both sectors were organized.
In what follows, we examine incentives for the export sector to get organized in
the presence of an already organized import-competing sector—for such an
outcome would take this economy from its initially distorted position to one of
free trade. In this context, members of the exportable group decide whether to
form a lobby or remain unorganized. To form the lobby, they face a fixed labor
cost. Nash interaction among group members is assumed in their contribution
decisions towards the provision of the fixed labor cost of lobby formation.
However, once the lobby is formed, it is assumed here that the lobby machinery
can enforce perfect coordination among the members of that group in the
collection of political contributions, i.e., given the symmetry of capital ownership
by members within a group, the lobby machinery can enforce collection of equal
amounts of political contributions from each capitalist in the sector.
Now, depending upon the magnitude of the fixed costs relative to the benefits of
lobby formation, there are three possibilities:
(1) The benefit to any one individual within the exportable lobby exceeds the
cost of forming the lobby. Here, contributing to the full financing of the fixed
P. Krishna and D. Mitra202
cost F is the only Nash equilibrium outcome among the group members, i.e.,
a lobby is always formed.
(2) Alternately, the cost of lobby formation exceeds the benefit to any one
individual but is less than the total benefit to the lobby. In this situation, there
are two possible Nash equilibrium outcomes—either there is no contribution
to the provision of the lobby or the fixed cost is fully financed. We assume
that pre-play communication can take place. For example, when capitalists in
an industry feel that they are going to benefit from forming a lobby, they start
communicating with each other—write letters, make phone calls, etc. Hence,
one can use some popular communication based refinements here. The better
equilibrium for the group (i.e., the lobby is formed) satisfies the conditions
for the three popular communication based refinements—coalition-proof
Nash, strong Nash and the Pareto-dominance refinement and hence, group
coordination becomes the likely equilibrium outcome.
(3) The cost of forming the lobby exceeds the benefit of lobby formation to the
group. Here, the Nash equilibrium outcome is obviously “not providing the
lobby”.
From the analysis of the above three cases, the conclusion that emerges is that a
lobby is formed when the total benefit exceeds the total fixed costs. Having
described the initial equilibrium that we focus on and having derived conditions
under which an (initially non-existent) export lobby may be formed, we proceed to
analyze the impact of unilateral trade reform on this initial equilibrium.
We are interested in how a unilateral tariff reduction by a large partner country
(leading to an improvement in the export price faced by the “small” home
country) may affect the initial equilibrium. In particular we are interested in how
this may affect the equilibrium structure of lobbies and finally on the equilibrium
structure of tariffs.
With truthful contributions, as in Grossman and Helpman (1994), any lobby
when formed will have to pay the government an amount that makes it indifferent
between treating that lobby as organized and treating it as unorganized, given the
contribution schedules of the other lobbies. As argued in Krishna andMitra (2003b),
the export lobby should compensate the government for the reduction in the import
lobby’s welfare due to its entry and for changes in overall social welfare. Netting
out these contributions in the calculation of the benefit from being organized, we are
able to show that with a pre-existing import competing lobby, the net benefit to
the exporting sector from the formation of an export lobby (gross of fixed costs) is
proportional to the sum of the deadweight losses created (relative to the free trade
level) in the importable and the exportable sectors by the equilibrium trade policies
Unilateralism in Trade Policy 203
that result when only the import competing sector is organized. The constant of
proportionality here is increasing in the government’s weight on welfare.
Let us now analyze the effect of the large partner country’s tariff liberalization,
that raises the world price of home’s exportable, on the structure of lobbies in the
home country. It needs to be noted first that we are able to show that a higher
world price of the exportable leads to a higher per unit (specific) export tax in
equilibrium when only the import-competing sector is organized. This is because
with a higher world price of the exportable, the gains from a given tax are higher to
the import lobby as consumers of the export good and as recipients of a fixed share
of the export tax revenues. Therefore, the first component of the effect of the
partner’s tariff liberalization is the change in the deadweight loss in the exportable
sector due to an increase in the absolute value of the tax. It is straightforward to see
that this effect is positive. The second component is the change in the absolute
value of the export tax for a given change in international export prices. We know
that this is positive as well. The first two effects are multiplicative, and therefore,
the resultant is positive. As we show in Krishna and Mitra (2003b), the third
component, which is just the effect of a higher world exportable price on the
deadweight loss at a given per-unit export tax, is also positive under the most
empirically plausible conditions on the curvature of the export supply function.
Thus, we establish our primary result that unilateral liberalization by a large
partner country within this framework will induce reciprocal liberalization.
Profiles of the net benefit from lobby formation and the per unit trade taxes as
functions of the world exportable price pp are illustrated in Figure 9.1. As can be
easily seen, once we have pp $ �p; we have free trade. It is also useful to interpret
the political-economy mechanism just stated in terms of how the welfare level
of the exportable group varies with the world price of the exportable differently
when this group is organized than when it is not. Figure 9.2 illustrates that an
increase in the world price of the exportable increases welfare of the exportable
group whether it is organized or unorganized. Thus, the welfare levels are shown
with positive slopes. Note that the welfare level of the exportable group when
organized (net of political contributions but gross of the fixed cost of lobby
formation) is higher than when it is not organized. Note also that the net benefit
NB (net of contributions but gross of fixed costs) from forming a lobby is
increasing in pp implies that the welfare locus when unorganized has flatter
slope—implying, in turn, with large enough fixed costs, some point of intersection
with the welfare locus when organized (net of both fixed costs and political
contributions). The price at which this takes place is, again, �p: Beyond this price,
the lobby is formed. Below it, it does not.
How might the large country benefit from such a political-economy dynamic?
Clearly, the (induced) movement of any single small country to free trade does not
P. Krishna and D. Mitra204
affectworld prices and its liberalization of its trade regime is of little consequence to
the large country. We, therefore, analyze here circumstances under which the large
country may nevertheless benefit from the induced reciprocity in its small trading
partners. Specifically, we now consider a large open economy trading with a
continuum of small open economies. These small open economies are identical to
Figure 9.1: Net benefits from lobby formation and trade taxes vs. world exportable price.
Unilateralism in Trade Policy 205
each other with respect to technology, endowments and tastes and preferences. We
assume also that the general structure of the “large” economy is similar to that of the
small open economies with which it trades (even though it may differ from them in
endowments and possibly the precise technologies used and in its exact prefer-
ences): its consumers have quasi-linear and additively separable utility as described
before. We assume that the structure of each of the small open economies is the
same as that of the small open economy in Section 9.2.1. We assume that the large
country exogenously cuts (sets) its import tariff and in each of the small open
economies, trade policy is set using political economy considerations as in
Grossman and Helpman (1994). From this analysis of a large country facing a large
number of small countries, we obtain the following two main results:
(a) For large enough reductions in tariffs by the large country, reciprocal
liberalization by the small open economies is likely.
Figure 9.2: Incentives to organize.
P. Krishna and D. Mitra206
(b) The large country’s optimal (i.e., welfare-maximizing) tariff is smaller when it
takes into account its effect on the incentives for lobby formation in its partner
countries thanwhen it takes the lobby structure in the partner countries as given.
Thus far, the argument has been developed in the context of a small trading
partner whose trade policy has been distorted due the exclusive initial presence of
an import-competing lobby. We next analyze the issue under other different initial
conditions:
(1) No lobbies are present in the small country initially.
(2) An export lobby rather than an import lobby is initially present.
In fact, we endogenize the initial conditions to the initial fixed costs, factor
endowments and world relative price. Therefore, at every point in time, which
lobby (lobbies) is (are) organized will be determined completely endogenously.
From our simulations, we now describe our findings pertaining to two important
cases. In our first case, for low values of the export price, we have a unique
equilibrium with just the import-competing sector organized. As the export price
rises, this initially continues to be the unique equilibrium. After pp rises even
further the export lobby gets formed and the unique equilibrium here involves the
formation of both lobbies (with free trade as the equilibrium trade policy) since the
net benefit from lobby formation make political organization feasible for both
sectors, each taking the other as organized. This scenario is consistent with what
we have focused on so far and illustrates our main argument. Note, however, that if
fixed costs were a bit higher in the import-competing sector, at low values of pp; theimport-competing sector (as well as the exportable sector) would not be organized.
If this were the initial condition, a reduction in tariffs by the large country would
now induce the import-competing lobby to form first—a change in economic
circumstances that would be welfare decreasing since agents in the small economy
will have moved from an efficient trade regime (with free trade) to an inefficient
one having incurred additionally the fixed costs of lobby formation. This may
appear damaging to the argument regarding the use of unilateralism to induce
reciprocity, but this is not the case since, of course, the argument is conditional on
some tariffs being imposed by the partner country in the first place. Also, with
further tariff reductions (i.e., increases in the world price, pp), the export lobby gets
formed andwe have free trade. Thus, with high enough tariff reductions by the large
country, free trade obtains in the partner country (even if the path to this is non-
monotonic and fixed costs of lobby organization have been incurred along theway).
Next, fixed costs of lobby organization for the export sector are assumed to be
lower than that in the importable sector. At low levels of pp; it is now the export
Unilateralism in Trade Policy 207
lobby that is organized while the import lobby is not. Trade policy is initially
distorted with export and import subsidies, an empirically nearly irrelevant case
but a clear theoretical possibility. Here, too, a high enough increase in the world
price of the exportable results in the formation of the import lobby with free trade
emerging as the policy outcome. While such an outcome would benefit the small
country (if the welfare gain from the move to the undistorted policy regime
outweighed the fixed costs of lobby formation), the large country would be faced
with a policy regime less favorable to it.
9.2.3. A model of leadership in trade policy negotiations
Almost every model of endogenous protection, while trying to incorporate
certain aspects of reality abstracts from others. Coates and Ludema argue that
firstly, “the trade policy process involves many political actors, none of whom
completely controls the outcome”. Secondly, they believe that “actions taken by
political actors inside a country may be unobservable to outsiders.” Incorporat-
ing these “real-world complications”, while taking a black-box approach to
lobbying (relative to the Grossman–Helpman framework), they look at a set
up where one large country (the leader) proposes to its partner a pair of tariffs
(its own and its partner’s). The partner then decides whether to accept or reject
the proposal. If the proposal is rejected by the partner, the two countries go
back to their initial tariffs. If the proposal is accepted by the partner country, it is
sent for political ratification within the country. The import competing sector in
the partner country can lobby to reduce the likelihood of ratification. If the
ratification does not take place, the large country unilaterally sets its tariff and
negotiations start once again.
Coates and Ludema show that lowering this unilateral tariff makes the proposal
of a given reciprocal tariff reduction more attractive to the partner country and at
the same time increases the probability of ratification. The higher attractiveness of
the proposal to the partner country government comes from the lower risk
associated with ratification success or failure in the sense that with a low tariff set
unilaterally by the large country the worse of the two outcomes is also not that
bad. This, effectively, is a situation of risk sharing between the two countries. The
lower probability of ratification failure arises from the fact that a lower unilateral
tariff by the large country means a lower relative price for the import-competing
good (a higher relative price for the exportable good) of the partner country for
any given tariff set by it, thereby making the protection situation relatively less
attractive to the import-competing lobby than before. In turn this will lead to less
intense lobbying by it against ratification.
P. Krishna and D. Mitra208
Thus, the equilibrium outcome in this model is one in which the large country
unilaterally liberalizes to increase the probability of reciprocation by its partner.
9.3. MODELS OF ENDOGENOUS UNILATERALISM
9.3.1. Unilateral commitment to free trade as a means of preventing
capital misallocation
Maggi and Rodriguez-Clare (1998) have an elegant and interesting political
economy explanation for the unilateral commitment to free trade agreements by
small countries. They formalize the frequently heard argument that free trade
agreements “provide a way for the government to credibly distance itself from the
domestic special-interest groups that lobby for protection”. More specifically, “the
idea is that, by committing to free trade, a government may be able to foreclose
political pressures at home”.
The setting in Maggi and Rodriguez-Clare is one in which owners of capital
first decide in which sector to invest and then those who invest in a particular
sector (the import-competing sector) lobby the government for protection. The
lobbying is modeled as a Nash bargaining game between the import-competing
lobby and the government over tariffs and political contributions. The lobby ends
up at least compensating the government for the deadweight losses purely
generated in the second stage. However, it may not compensate the government
for the welfare loss through the intersectoral misallocation of capital in the first
stage in the expectation of protection in the second stage. In such a situation, it is
possible that a government may exercise its option, if available, of committing to a
free trade agreement in a prior (to stage one) stage zero. Such a situation is one in
which, in the absence of the agreement, the welfare loss from the resource
misallocation in the first stage is valued more by the government than its gain from
sharing the redistributed surplus in the second stage.
9.3.2. Unilateral commitment to free trade as a means of preventing
wasteful political (organizational) activity
The Maggi–Rodriguez-Clare framework demands a government with a long
enough horizon as intersectoral capital mobility is a fairly long-run phenomenon.
Such an assumption is perfectly valid when the focus is on developed countries
that have stable governments. However, in the recent past, quite a few developing
countries have joined or have expressed a desire to join the GATT/WTO. In such
countries, governments are generally weak and often do not last long. In such
Unilateralism in Trade Policy 209
situations, they could hardly be expected to care about long-term problems such as
capital misallocation and thus capital mobility may not be an aspect one would
like to focus on. With the frequent entry and exit of parties into and from power,
lobbies need to constantly incur costs build new relationships.
In this context, Mitra (2002) builds on the Maggi–Rodriguez-Clare version of
the Grossman–Helpman framework, augmenting it with the decision to incur
fixed costs (build relationships with politicians in power and/or to form a lobby)
prior to the actual lobbying, but, importantly, not providing room for any capital-
mobility. However, the main result of the Maggi–Rodriguez-Clare model goes
through even in this newly modified set up. This is the result that generally
governments with low bargaining power with respect to domestic lobbies are the
ones that want to precommit to free trade agreements.
Thus, there is a general point to be made here, which is that the precommitment
to a free trade agreement does not have to be driven specifically by the possibility
of capital misallocation alone (or solely by the possible incurring of organizational
costs) arising in the expectation of protection. It is applicable to any kind of
resource costs (including, for example, costs of political organization) incurred
prior to lobbying through actions taken in the expectation of successful lobbying
in the next stage. In this respect, the paper by Mitra and the one by Maggi and
Rodriguez-Clare are complementary.
9.4. CONCLUDING REMARKS: THE CURRENT STATE OF THE
LITERATURE AND ISSUES FOR FUTURE RESEARCH
The debate over the merits of unilateralism versus reciprocity is a long standing
one. Of special interest to us is the role of unilateralism in trade policy in inducing
reciprocity from trading partners. While much of the literature has considered the
merits of the two policy stances of unilateralism and reciprocity independently
of each other, we focus on the possible causal connection between the two.
Can unilateral trade liberalization by one country lead to reciprocal liberalization
by its partner in the absence of negotiations between the two countries? Krishna
and Mitra (2003a,b) have considered this issue analytically in the context of two
separate political economy models of trade policy determination. While one is
based on majority voting, the other is based on lobbying through political
contributions to the incumbent government when the formation of lobbies is
endogenous. In this survey chapter, we provide a non-technical, descriptive and
intuitive treatment of these two models.
While the aspects of voting and lobbying are rigorously modeled with full
microfoundations in our two papers, Coates and Ludema, by usefully abstracting
P. Krishna and D. Mitra210
from the microfoundations of lobbying, focus on other real-world complications
of the political process. They incorporate aspects like leadership in negotiations,
domestic political ratification of international agreements and the imperfect
observability of the political process in foreign countries into their analysis. They
also predict a “reciprocated unilateralism” in their setting which is considerably
more “cooperative” than those we considered. Given that the real world is a
combination of cooperative and non-cooperative interactions, we believe that our
research and that of Coates and Ludema are complementary.
In addition to examining the scope for unilateral policies to induce reciprocity
in partner countries, we have also covered briefly the analysis provided in the
literature (by Maggi and Rodriguez-Clare (1998) and then by Mitra (2002)) of
other contexts in which unilateral trade liberalization may be undertaken. It goes
without saying that reciprocity by partners of the sort we have discussed may
follow even in these contexts.
Finally, given the current state of the literature on unilateral trade liberalization,
what do we think are the open questions and issues for future research in this area?
In our work in Krishna and Mitra (2003b) as well as in Coates and Ludema, the
political economy of the unilaterally liberalizing country has not been rigorously
modeled. What is fully modeled is the political economy response of the recipro-
cating country. Thefinal step in understanding “reciprocated unilateralism” has to be
the construction of a completely closed model with the full modeling of political
economy forces in both countries. In fact, our median-voter model in Krishna and
Mitra (2003a) is completely closed and free of black boxes, even though somewould
consider amajority-voting-based story to be less realistic thanonebasedon lobbying.
The focus of this chapter has been on unilateralism. We look at what the trade
literature, especially the strand that employs political economy models, has to
say about the desirability and feasibility of unilateral trade liberalization. While
classical trade theory clearly has a lot to say about desirability, the feasibility issue
has been studied in some detail by Maggi and Rodriguez-Clare (1998) and then by
Mitra (2002).
Another issue for future research are the reasons behind the recent unilateral
trade reforms by many countries. While there is a large and sophisticated literature
trying to explain the existence of protection, there is no satisfactory work on why
it is disappearing. Models need to map the actual changing conditions in
liberalizing countries to key assumptions in models to generate an equilibrium
outcome of trade liberalization.
Finally, we also need to investigate why unilateral reforms take place in stages
and why there are so many rounds of tariff cuts. A model of learning in which
information gets uncovered over time might be promising. Also, reforms in stages
might be a way of working around domestic political constraints.
Unilateralism in Trade Policy 211
REFERENCES
Bagwell, K. and Staiger, R. (1999). An economic theory of GATT. American Economic
Review, 89, 215–248.
Bhagwati, J. (1971). “The generalized theory of distortions and welfare,” in Trade Balance
of Payments and Growth: Papers in International Economics in Honor of
Charles Kindleberger, J. N. Bhagwati, R. W. Jones, R. A. Mundell and J. Vanek
(eds.), Amsterdam: North-Holland.
Bhagwati, J. (1990). “Aggressive unilateralism,” in Aggressive Unilateralism, J. Bhagwati
and H. Patrick (eds.), Ann Arbor, MI: University of Michigan Press.
Bhagwati, J. (2002). Free Trade Today, Princeton, NJ: Princeton University Press.
Coates, D. and Ludema, R. (2001). A theory of trade policy leadership. Journal of
Development Economics, 65(1), 1–29.
Fernandez, R. and Rodrik, D. (1991). Resistance to reform: status-quo bias in the presence
of individual-specific uncertainty. American Economic Review, 81(5), 1146–1154.
Grossman, G. and Helpman, E. (1994). Protection for sale. American Economic Review, 84,
833–850.
Grossman, G. and Helpman, E. (1995). Trade wars and trade talks. Journal of Political
Economy, 103, 675–708.
Hillman, A. and Moser, P. (1996). “Trade liberalization as politically optimal exchange
of market access,” in The New Transatlantic Economy, M. B. Canzoneri, V. Grilli and
W. J. Ethier (eds.), Cambridge, UK: Cambridge University Press, 295–316.
Johnson, H. (1953). Optimum tariffs and retaliation. Review of Economic Studies, 21,
142–153.
Johnson, H. G. (1965). “Optimal trade intervention in the presence of domestic distortions,”
in Trade, Growth and the Balance of Payments, Caves, Johnson and Kenen (eds.).
Krishna, P. and Mitra, D. (2003a). Reciprocated unilateralism: a median-voter approach,
mimeo, Brown University and Syracuse University.
Krishna, P. and Mitra, D. (2003b). Reciprocated unilateralism in trade policy. Journal of
International Economics, in preparation.
Maggi, G. and Rodriguez-Clare, A. (1998). The value of trade agreements in the presence
of political pressures. Journal of Political Economy, 106(3), 574–601.
Mayer, W. (1981). Theoretical considerations on negotiated tariff adjustments. Oxford
Economic Papers, 33, 135–143.
Mayer, W. (1984a). The political economy of tariff agreements. Schriften des Vereins fur
Socialpolitik, 148, 423–437.
Mayer, W. (1984b). Endogenous tariff formation. American Economic Review.
Mitra, D. (1999). Endogenous lobby formation and endogenous protection: a long run
model of trade policy determination. American Economic Review.
Mitra, D. (2002). Endogenous political organization and the value of trade agreements.
Journal of International Economics, 57, 473–485.
P. Krishna and D. Mitra212
CHAPTER 10
Trade Creation and Residual Quota
Protection in a Free Trade Area
with Domestic Monopoly
DAVID H. FELDMANa,* and MARTIN RICHARDSONb
aDepartment of Economics, College of William and Mary, Williamsburg, VA 23187-8795, USAbFaculty of Economics and Commerce, School of Economics, The Australian National University,
Crisp Building 026, Australian National University, Canberra ACT 0200, Australia
Abstract
This chapter explores the consequences of discriminatory liberalization in the
presence of quota-protected single-firm industries. We show that a free trade area
may not lead to any trade creation, though it must be welfare improving in the
usual sense of increasing real national income. We use a simple political economy
model to show that profit must have a higher weight to the policy maker than quota
revenue for an free trade area (FTA) to dominate the initial quota-distorted
equilibrium. Lastly, moving into an FTA has an ambiguous effect on the policy
maker’s incentives to liberalize the residual quota on non-members.
Keywords: Free trade area, quota, domestic monopoly
JEL classifications: F13, F15
*Corresponding author.
E-mail address: [email protected]
10.1. INTRODUCTION
This chapter explores the consequences of discriminatory trade liberalization in
the presence of quota-protected single-firm industries. In recent years, trade
liberalization among developing nations has occurred increasingly outside of the
multi-lateral, GATT-based, negotiating rounds. Instead we have discriminatory
“mini-lateral” arrangements like NAFTA, MERCOSUR, the Andean Pact, or
ASEAN. At the same time, protection often takes the form of quantitative
restraints and other non-tariff barriers.1 Lastly, the single-firm industry is a
common occurrence in developing economies both currently and historically.
Small domestic markets create a predisposition for dominant-firm local
industries. As Tybout (2000, p. 29) notes, the log of GDP alone explains two-
thirds of the cross-country variation in measures of industrial concentration.
Concentration is stronger in intermediate goods and durable consumer goods,
many of which compete with imports.2
Policy often has complemented geography in creating highly concentrated
domestic industries in the developing world. Many public and private enterprises
have received protection in the form of non-tariff barriers (NTBs) coupled with
special privileges including monopolies over domestic production and distri-
bution. And as Krueger (1993, pp. 20–21) notes, rationing of imported
intermediate inputs coupled with licensing of new investment effectively
guarantee high market share for existing quota-protected domestic firms.
Despite recent moves toward liberalization, the industrial sector remains
relatively protected in the typical developing economy (Ng, 1996), and many
highly-concentrated industries persist. Remaining non-tariff barriers often protect
the most highly concentrated domestic sectors.3 Privatization itself sometimes
1 A sampling of the political-economy literature on the choice between tariffs and quotas includes
Young and Anderson (1980, 1982), Cassing and Hillman (1985), and Kaempfer et al. (1989). Vousden
(1990) provides a good summary. See also Sweeney et al. (1977) for the positive economics of the
choice between tariffs and quotas under monopoly.2 Whitworth (1980), for example, examined the industrial structure of Tanzania in the mid-1970s. Of
31 domestic industries employing more than 50 people, 16 had but a single firm. Some of these
produced goods with a high ratio of transport cost to value (cement, boxes, tires), but others faced more
significant potential competition from imports (cigarettes, aluminum household goods, steel
manufactures, and farm implements). More recently, quota-protected single-firm industries appear in
the firm-level Moroccan data set used by Dutz (1996), and in newly industrializing East Asian
economies examined by De Melo and Roland-Holst (1991, p. 290). Of 2260 Korean sectors surveyed
by them in the early 1980s, in 533 cases more than 80% of domestic output was controlled by a single
firm.3 In Africa, these often include sugar, beer, soft drinks, cement, fertilizer, and steel products. (World
Bank, 1994, ch. 3).
D.H. Feldman and M. Richardson214
accentuates concentration.4 The single-firm case also is a useful simplification if
domestic industries are more highly concentrated than those in the preferred
partner.5
Meade (1955, pp. 96–99) was the first to examine discriminatory liberalization
in the presence of quotas. He showed that a small nation whose quota-protected
domestic markets are competitive must gain from a free trade area (FTA) with a
preferred partner whose supply price is less than the quota-distorted domestic
price. All increased domestic consumption comes from trade creation, and trade
diversion is not present.6
Our first task is to replicate Meade’s comparative statics for the case where the
domestic industry is composed of a single firm. We show that an FTA may not
lead to any trade creation, though it must be welfare improving. In the usual
Vinerian framework, trade diversion creates a tension between regional and
multilateral trading arrangements. In this monopoly model no trade diversion
occurs, so one common defect of regionalism is not present. While this result is
not counterintuitive to our knowledge it has never been noted anywhere in the
long literature on trade policy in the presence of domestic monopoly.
The single-firm case also affects the political economy of integration. In Section
10.3, we use a simple political-support maximization model to explore the
desirability of FTA formation and its probable consequences on the politically
optimal level of residual quota protection. In our endogenous protection model the
policymaker finds the optimal trade-off between political support from policy
surpluses (composed of enterprise profits and quota rents) and higher prices that
antagonize consumers. Like a quota liberalization, an FTA that lowers domestic
4 Lustig (1992) gives a number of examples from Mexico. After privatization, a single firm controlled
100% of the chloric acid industry, 96% of copper output, and 67% of the cement industry. The top four
paper producers controlled over 60% of output. Many other markets in Mexico are highly concentrated.
As Ortiz (1991) noted, price controls led to production increases in many Mexican industries. This is
consistent with the behavior of single firm industries (see Helpman, 1987).5 While single-firm industries are common in many LDCs, the qualitative features of our analysis are
likely to hold in more oligopolistic structures too. The key feature in our analysis distinguishing
monopoly from competition is that an FTA may not lead to trade creation under quota protection in the
former case. In the competitive case, high cost local production is displaced by imports from the
partner. Under monopoly, increased domestic consumption may be made up by the domestic
monopolist. As in Krishna (1989), the effects of a quota on a competitive domestic industry are present
in the monopoly case but the latter also contains effects due to elasticity of demand changes; and the
effects in the monopoly case are present in oligopoly along with other strategic interaction effects. The
exact modelling of oligopoly will determine the nature of any further strategic interactions amongst
domestic firms, but the effects we identify here will still be relevant in that setting.6 Goodman and Kaempfer (1993) formalize Meade’s proposition and extend it to the large-country
case.
Trade Creation and Residual Quota Protection in a Free Trade Area 215
price will reduce policy surplus. Thus a proposed FTA could be desirable only if
the political support obtained from enterprise profit and quota revenue at the
partner’s price were sufficiently higher than with the price-equivalent quota.7 We
show that a necessary condition for an FTA to dominate the initial quota-distorted
equilibrium is that profit must have a higher weight to the policymaker than quota
revenue. This follows from the fact that an FTA raises profit, but lowers total
surplus (profit plus revenue) at any given partner price, relative to the price-
equivalent quota.
The level of residual protection within an FTA is another contentious, but
sparsely modeled issue. Richardson (1993) uses political-support maximization to
show why residual tariff protection of a competitive industry likely would fall.
Once in an FTA the initial non-discriminatory tariff becomes partially redundant
and the policymaker can reverse all trade diversion by reducing the tariff on non-
members. Unlike a tariff, quota protection does not become redundant within an
FTA. We show that a policymaker in these circumstances may or may not have an
incentive to liberalize the initial quota. In the case examined here, discriminatory
arrangements may not provide effective stepping stones to further liberalization,
though they need not hurt third parties who might retaliate in ways that erode the
multilateral system. Section 10.4 formalizes these arguments.
10.2. SINGLE DOMESTIC FIRM
This section explores the comparative statics of FTA formation where the
domestic single-firm industry is protected by a fixed quota. We presume the
domestic economy is small both on world markets and within the proposed FTA.
The rest of the world (ROW) is the low cost producer and supplies the good
perfectly elastically at a price Pw: In Figure 10.1, D is domestic demand and Dd is
the net-of-quota demand facing the domestic firm, i.e.,D2 Q: Each demand curve
has an associated marginal revenue curve, and the domestic firm faces increasing
marginal costs. The quota allotment is filled initially by imports from ROW.
Suppose for now that the quota is a world-wide quota and that the rights are
sold. In a non-competitive setting, auctions may not be able to capture quota rents
in their entirety. If there is market power on the buyers’ or sellers’ side of product
markets or license markets, license prices may fall short of the potential rent.8 The
single-firm industry examined here does not exhibit this problem. If the firm were
given the quota rights it would import the full allotment from the low cost source
7 Price-equivalence here refers to the quota that would decrease domestic price by the same amount as
the FTA.8 See Krishna (1993) for a thorough survey of this literature.
D.H. Feldman and M. Richardson216
and earn the quota rent ð5; 6; 7; 8Þ as profit.9With full information, the government
could set the license price exogenously and convert quota rent into revenue. Under
a fixed quota the domestic market is non-contestable so the firm chooses its output
level ðQdÞ to equate marginal revenue from its residual demand curve with
marginal cost. The quota-distorted domestic price is Pd: Deadweight losses areðPw; 12; 11; 5Þ and ð7; 8; 16Þ: Quota revenues equal ð5; 6; 7; 8Þ:
10.2.1. An FTA may not lead to any trade creation
Suppose now that the preferred partner’s supply price is Pp: Given the above
assumptions the domestic market is now contestable at the partner’s supply price.
Thus MR ¼ Pp for the domestic firm and its output rises along MC from point 11
to point 2. The quota allotment continues to come in from the low cost source at
Figure 10.1: Single-firm industry.
9 If the quota rights price is set exogenously by the authorities, a monopoly firm whose output decisions
affect domestic price may purchase the entire quota but exercise only a portion of it. This “water-in-the-
quota” is analyzed by Hillman et al. (1980). Water-in-the-quota becomes more likely if the domestic
firm is “large” in the sense that its free-trade share of the market would be high. Consider a domestic
firm whose MC curve intersects MR below Pw: Such a firm would supply more than half the free-trade
consumption level. Here a competitive auction would not be feasible since the social marginal value of
imports is less than the marginal value to the monopoly of destroying a license. Nevertheless, the full
quota revenue could be realized by the government through a number of techniques, including price
ceilings or direct government imports.
Trade Creation and Residual Quota Protection in a Free Trade Area 217
price Pw; and quota revenues equal ð1; 2; 3; 4Þ: In this example no trade is created
since the increase in domestic consumption is supplied by the domestic firm. The
threat posed by goods from the preferred partner changes the domestic firm’s
behavior, though no imports from that source actually enter the domestic
market.10
The price Pp in Figure 10.1 is a critical margin, as it is the price at which MC
intersects Dd: For preferred partner’s prices below Pp imports equal to the
horizontal distance between Dd and MC do enter from the partner and trade is
created, though some of the increased domestic consumption may be supplied by
higher domestic output. For partner’s prices between Pp and Pd the domestic firm
produces along Dd between points 2 and 6. Though the potential entry of goods
from the preferred partner sets the domestic price, the residual demand facing the
domestic firm is Dd: The only imports come in from the low cost ROW producer
who uses the full quota allotment.
Suppose we replace our working assumption that the quota is a world-wide
quantity constraint with a country-by-country quota under which the preferred
partner and the ROW producer each have a share of the initial quota. For
simplicity, we will assume the shares are 50%, but the result is easily generalized.
Assume also that the shares are not transferable. This scenario requires either a
country-by-country auction or (more reasonably) a transfer of the quota rents to
producers abroad, as in the US sugar quota disbursement system. The residual
demand faced by the domestic firm remains Dd so the initial domestic price is still
Pd: The partner and the ROW producer each sell 0:5Q in the domestic market. If
the partner’s price and the ROW price differ then the value of the country-specific
quotas would differ as well.
Relaxing the quota on the preferred partner shifts Dd to the right by 0:5Q in
addition to making the domestic market contestable at the partner’s supply price.
As in the previous example, if the partner’s price is Pp the domestic firm will
produce at point 2 on its marginal cost curve. Thus imports from the preferred
partner in the FTA would exactly equal its initial quota allotment, i.e., there still is
no trade creation. For partner’s prices above Pp imports would fall in comparison
to the world-wide quota case. Thus with a country-by-country quota an FTA may
lead to trade destruction though, as we show below, without the usual negative
welfare consequences.
10 The partner’s price should properly be viewed as a relative price of the firm’s output. The partner
presumably levies a quota or tariff on some other good imported from the home economy. Relaxing that
barrier reduces the relative price of the good considered here.
D.H. Feldman and M. Richardson218
10.2.2. Welfare effects
In the competitive case, trade creation is assured and FTA formation proves
unambiguously welfare enhancing. In the single firm case, trade creation is not
assured. The FTA may worsen domestic overproduction relative to the
competitive outcome. Nevertheless, the aggregate welfare effects remain
unambiguously positive. The FTA guarantees domestic market contestability
since exports from the preferred partner can enter in unlimited amounts at the
partner’s supply price. Although the FTA has the potential to move the economy
further from the first best (free trade) equilibrium, the contestability gain proves
dominant. Thus we have a case in which a second best policy response (the FTA,
which is a second distortion), when combined with the initial distortion (the
quota), must raise welfare.
Using Figure 10.1, consider again the case in which the preferred partner’s
supply price is Pp and the auctioned quota rights sell for their full value. The
consumption gain is ðPd; 7; 3;PpÞ: The firm’s profit falls by ðPd; 6; 9;PpÞ2
ð9; 2; 11Þ; while quota revenue declines by ð7; 6; 9; 10Þ: The net change in
welfare is thus ð7; 10; 3Þ þ ð9; 2; 11Þ: Alternatively, we can measure the gain in
terms of deadweight loss before and after the FTA. The fall in consumption
deadweight loss ð1; 2; 6; 5Þ minus the rise in production deadweight loss
ð1; 2; 11; 5Þ gives the net gain ð6; 2; 11Þ ¼ ð7; 10; 3Þ þ ð9; 2; 11Þ:Similar logic shows a welfare gain for all partner’s prices below Pd: This is
clear if the FTA leads the domestic firm to produce less as this corrects the
over-production distortion of the quota as well as yielding consumption gains.
To see that it also holds even if domestic output increases, consider some
partner’s price P†, Pd to which the domestic firm responds by producing
some quantity Q†. Qd: As the domestic price falls from Pd to P† there is a
transfer from the domestic firm to consumers of ðPd 2 P†ÞQd as well as an
increase in the domestic firm’s profits corresponding to the area above the MC
curve but below P†; between Q† and Qd: On top of this gain, there are further
gains to consumers represented by the consumer surplus triangle on Dd below
Pd and above P†:Indeed, for partner prices below the quota-distorted price under perfect
competition (Pp in Figure 10.1) the FTA yields larger gains in the single firm case
than in the corresponding competitive case. With a single-firm industry, the FTA
induces an additional gain due to market contestability. Domestic welfare is thus a
monotonic decreasing function of the partner’s price.11
11 The function is not concave to the origin. For partner’s prices between Pd and Pp; welfare increases
at a decreasing rate, while prices progressively lower than Pp yield gains that increase at the margin.
Trade Creation and Residual Quota Protection in a Free Trade Area 219
10.3. THE POLITICAL ECONOMY OF FTA FORMATION
We have shown that an FTA must increase aggregate economic welfare in this
setting, and that the lower the partner price the greater the rise in welfare. This
does not imply that an FTA also must be politically desirable. Nor does it suggest
that the FTA becomes more appealing the lower the partner’s price. In this section
we explore the political economy of FTA formation between a larger partner and a
smaller home country in which quota protected single-firm industries are an
important part of the economy.
We evaluate also whether FTA formation encourages or discourages further
liberalization of the initial quota. Richardson (1993) has studied how FTA
formation affects residual tariff protection of a small competitive industry. In that
setting, the portion of the tariff equal to the difference between the tariff-distorted
domestic price and the partner’s price becomes redundant within the FTA. The
policymaker then can reduce the external tariff by an amount epsilon larger than
the redundancy. This switches imports back to the rest of the world and reverses
the initial trade diversion.12 Richardson (1993) shows that reducing the residual
tariff must increase political support as long as the policymaker welfare function
gives some positive weight (no matter how small) to consumer surplus and tariff
revenue relative to producer surplus. With our single-firm industry, the initial
quota does not become redundant within the FTA, so the liberalization incentives
at work in Richardson’s model are not present here.
We model the political process using a policymaker welfare function. This
approach is useful in understanding government behavior in many developing
countries in which the state itself is a strong player and the multiplicity of private
interests (civil society) that predominate in more pluralistic settings are weaker
followers. Formally, we assume policymaker welfare depends on policy surplus
(firm profits and quota revenues) and domestic price
W ¼ w½rðp;RÞ;P� ð1Þ
where r refers to policy surplus, the components of which are industry profit ðpÞand quota revenue ðRÞ; and P ¼ Pd 2 Pw is the quota-induced wedge between
domestic and world prices. Policy-induced increases in domestic price antagonize
consumers and reduce policymaker welfare, while added profits and quota rents
raise it, or wr ¼ ð›W=›RÞ . 0 and wp ¼ ð›W=›PÞ , 0: The politically optimal
quota equates at the margin the policymaker’s willingness to substitute between
12 Feldman (1993a) shows why redundant tariff protection may persist in a political equilibrium given
world price uncertainty.
D.H. Feldman and M. Richardson220
policy surplus and domestic price with the transformation relation between
the two.13
The first order condition for an optimum is,
2wp
wr
¼ rpdp
dPþ rR
dR
dP. 0 ð2Þ
where rp and rR are the weights on profit and revenue in the utility function. Profit
is monotonic increasing in price between the world price (no quota barrier) and
the monopoly price (a zero quota). As the government’s first order condition
(Equation 2) demonstrates, the politically efficient quota is unlikely to yield
maximum quota revenues. It may be so restrictive that dR=dP , 0: This could
occur if dp=dP is sufficiently high, and/or if the weight on profit at the margin
sufficiently exceeds the weight on quota revenue.
In general, quota revenues are given by
R ¼ PQ ð3Þ
where Q is the quota amount. Domestic price is a function of the quota, so
dR
dQ¼ Pþ Q
dP
dQð4Þ
The firm maximizes profits given the residual demand curve it faces,
P ¼ P½QðPdÞ2 Q�2 cðqÞ ð5Þ
where QðPÞ is total quantity demanded, q ¼ QðPdÞ2 Q is the firm’s output and
cðqÞ is the cost function. As the interesting case here is where the country would be
an importer of this good, we assume henceforth (as in Figure 10.1) that c0ð0Þ $
Pw; which ensures that the domestic firm’s marginal cost exceeds the world price
for all levels of domestic output. We then have the standard first order condition
for a monopoly,
dP
dq¼ qP
0dðqþ QÞ þ Pdðqþ QÞ2 c
0ðqÞ ¼ 0 ð6Þ
13 While we could specify a particular form for the government’s objective function—writing down an
explicit model of the link between rents and profits and lobbying, for example—we have left it in this
form so as not to obscure the generality of the analysis. Endogenising the policy maker’s maximand,
while an interesting task, is not the focus of this chapter.
Trade Creation and Residual Quota Protection in a Free Trade Area 221
From this,
dq
dQ¼ 2
ðP0d þ qP00
dÞ
ð2P0d þ qP00
d 2 c00Þð7Þ
and hence,
dP
dQ¼ P
0d 1þ
dq
dQ
!
¼ P0d
P0d 2 c00
2P0d þ qP00
d 2 c00
!
, 0 ð8Þ
where the sign in Equation 8 follows from the firm’s second-order condition.
From Equations 4 and 8 the effect on policy surplus of relaxing the quota is
ambiguous. For domestic prices above the revenue maximizing level, quota
revenue falls more rapidly than profit rises, so policy surplus may decline. This
policy surplus relationship is the binding constraint on the policymaker’s
welfare.
10.3.1. FTA is politically unattractive when profits and revenues
are equally valued
This endogenous policy model has been used before to examine the choice
between tariffs and quotas.14 If the policymaker places equal value on a unit of
enterprise profit or a unit of tariff/quota revenue, then quotas dominate tariffs since
the sum of profits and revenues under a quota is always as large or larger than
under the price-equivalent tariff. This equal weighing of profit and revenue is
consistent with a regime in which individuals (groups) who have claims to the
state’s revenue also have claims to enterprise profits. As the weight on revenue
approaches zero, the model collapses into the Cassing–Hillman (1985) case in
which tariffs politically dominate quotas.
We begin the formal analysis by examining the effects of an FTA on
enterprise profit and quota revenue separately, and conclude by showing that a
policymaker that weighs revenue and profits equally might never choose to form
an FTA in this setting. The political desirability of FTA formation depends on
how an FTA affects the trade-off available to the policymaker between policy
surplus and domestic price. Thus we must first compare how an FTA that lowers
14 See Cassing and Hillman (1985) and Feldman (1993b).
D.H. Feldman and M. Richardson222
domestic price affects profit and revenue relative to a price-equivalent quota
liberalization.
Let us suppose that the surplus function can be written as,
rðp;RÞ ¼ dpþ ð12 dÞR ð9Þ
where d [ ½0; 1� is the weight placed on profits versus quota rents. Then, in the
absence an FTA, a government choosing Q to maximise Equation 1 obtains, from
Equation 2,
0 , 2wp
wr
¼ ddp
dPþ ð12 dÞ
dR
dPð10Þ
The domestic price that follows from Qp; the optimal quota implicit in Equation
10, we shall denote Pp
d; and the level of welfare that obtains we shall denote Wp:Suppose that this government now forms an FTA and faces a partner price of
Pp , Pp
d: Denote the level of welfare that follows if the quota is unchanged as
Wpp: Now, the domestic price will fall to Pd ¼ Pp so Q ¼ DðPpÞ2 Q ; Qp: Thusp ¼ PpQp 2 cðQpÞ and R ¼ ðPp 2 PwÞQ; so
dp
dQ¼ ðPp 2 c
0ÞdQp
dQ¼ 2ðPp 2 c
0Þ ð11Þ
and
dr
dQ¼ d
dp
dQþ ð12 dÞðPp 2 PwÞ ¼ ð12 dÞðPp 2 PwÞ2 dðPp 2 c
0Þ ð12Þ
Thus in the case where revenues and profits are valued equally ðd ¼ 12Þ we get
dr
dQ¼ 2
1
2ðPw 2 c
0Þ ð13Þ
So long as MC exceeds the world price, our maintained assumption, this
government would wish to expand the quota. A small expansion would have no
effect on domestic price (unchanged at Pp) and thus no effect on consumers, but
would increase rðp;RÞ: The liberalization thus raises government welfare
monotonically above Wpp; but eventually the quota will be expanded to the
level that would have sustained a domestic price of Pp in the absence of an FTA.
At this point the level of government welfare has risen to that which would have
prevailed had they chosen the quota that yielded Pp without an FTA. By definition,
Trade Creation and Residual Quota Protection in a Free Trade Area 223
this is less than Wp (as the initial quota was optimal) so, overall, the government
can be no better off in the FTA than without it.15
Intuitively, the government sets its optimal quota initially to trade-off not only
policy surplus against consumer surplus but also revenues against domestic
profits. In Figure 10.1, if Q is the optimal quota then the marginal profit from
another unit of domestic production is given by height ð6; 11Þ and marginal quota
rent from a one-unit quota expansion is ð6; 5Þ: When an FTA is formed and the
domestic price falls, this latter trade-off is disturbed: the impact of a lower price on
the marginal profit ðPd 2 c0Þ is greater than that on the marginal quota rent ðPd 2
PwÞ: Equation 12 shows that the optimal trade-off equates the d-weightedmarginal profit with the ð12 dÞ-weighted marginal quota rent so marginal profits
must rise relative to marginal rents and this leads to an expansion of the quota. In
Figure 10.1, if the quota is unchanged after the FTA is formed then marginal profit
falls to zero (as the domestic firm produces at point 2 where Pp ¼ MC) while the
marginal quota rent is given by ð2; 1Þ: Expanding the quota pushes the domestic
firm back down its MC curve and raises the marginal profit accordingly. That
expansion raises policy surplus continuously until the quota is reached that would
have sustained Pp as a domestic price in the absence of an FTA; that is, in Figure
10.1, until a quota of ð3; 14Þ is reached (at which point policy surplus has
expanded from ð2; 12;PpÞ þ ð1; 2; 3; 4Þ to ð14; 16; 12;PpÞ þ ð3; 4; 15; 14Þ: At thispoint total welfare is the same as that which the government could have achieved
by setting a quota of ð3; 14Þ in the first place with out FTA: by revealed preferencethis cannot be as desirable as Q:
So far we have taken the usual approach to an FTA and assumed that the
partner’s price operates as a ceiling price only in the domestic country. Richardson
(1995) notes that it might also be considered a price floor in that the domestic firm
can always sell in the partner at that price. If this were the case then, in the analysis
above, the government would wish to expand the quota up to the full quantity
demanded—DðPpÞ—once the FTA is formed! The reason is that the domestic firm
would produce at point 2 in Figure 10.1 but sell its entire output in the partner
country at price Pp and the total volume of domestic consumption—out to point
3—would be imported from the rest of the world, yielding quota rents of
15 The mathematics in Equations 11–13 holds for partner prices between pd and pp in Figure 10.2
(outcomes without trade creation). The logic, however, extends to the trade creation case as well. Once
in an FTA the firm’s marginal revenue curve becomes perfectly elastic at the partner’s price until its
output equals residual demand. Increasing the quota on non-member imports would have no effect on
domestic price until the quota equaled the amount that, in the absence of an FTA, would have driven
price down to the partner’s price. At best, a policymaker that values profit and revenue equally always
faces the original trade-off between policy surplus and domestic price.
D.H. Feldman and M. Richardson224
ð3; 4;Pw;PpÞ: In this case the FTA is clearly desirable even for a government that
weighs revenue and profit equally.
10.3.2. FTA can only be politically attractive if profits are more
highly valued than revenues
Returning to the case where the FTA provides a price ceiling and not a floor, is it
possible that an FTA could be attractive to a government that did not weigh
revenue and profits equally? This could only be the case if an FTA can offer a
trade-off that is not otherwise available, and in this section we identify exactly
such a tradeoff. Consider Figure 10.1 and suppose that the optimal quota, in the
absence of FTA, was ð3; 14Þ: This yields a domestic price of Pp; correspondingconsumer surplus, domestic profits of ð12; 16; 14;PpÞ and quota rents of
ð3; 4; 15; 14Þ: If the home economy joins an FTA with a partner whose price is
Pp and the initial quota is maintained, then the FTA will have no effect. If the
quota is tightened, however, domestic price will not change but domestic output
will increase along the MC curve potentially as far as point 2. With a quota of Q;for instance, the FTA gives profits of ð2; 12;PpÞ and rents of ð1; 2; 3; 4Þ; as we havenoted earlier. The important point here is that profits are higher under the price-
equivalent FTA and appropriate quota than under a quota alone, even though the
sum of revenue and profits is lower. Accordingly, one would anticipate that
joining an FTA might be more attractive to a government that values profits over
revenues. Of course, the FTA will reduce profit if the partner’s price is too low,
which also suggests that the FTA is more likely to be attractive when the partner’s
price is “close” to the pre-FTA domestic price.
To answer this question more generally we must compare welfare levels
between two discretely different regimes: an optimally chosen quota alone, and
an FTA with an optimally adjusted quota. This is intractable even in this simple
model so we turn to some numerical simulations to illustrate the intuitive results
suggested above. We simulate a linear version of the model above in which
Pd ¼ a2 bQ; w½r;P� ¼ bCSðPdÞ þ ð12 bÞr; rðp;RÞ ¼ dpþ ð12 dÞR and
MCðqÞ ¼ aþ gq for constants a; b; b; d; a and g: We confine our attention
to cases in which a quota is initially preferred to a tariff (in the absence of an
FTA). In this setting we can show that the optimal quota when an FTA is joined
(given that domestic production remains positive) is given by
QFTA
¼1
bdg½dðgþ bÞða2 PpÞ þ bð12 dÞðPp 2 PwÞ2 bdða2 aÞ� ð14Þ
Trade Creation and Residual Quota Protection in a Free Trade Area 225
We can also calculate the maximum value of this FTA quota such that
the domestic price is exactly Pp—this is the equivalent of the quota ð3; 14Þ inFigure 10.1:
Qmax
¼1
bðgþ bÞ½ðgþ bÞða2 PpÞ þ bða2 PpÞ� ð15Þ
Following our earlier reasoning, if QFTA. Qmax it must be true that the FTA is
politically undesirable: the welfare level achieved at Qmax was attainable in the
absence of the FTA by setting Q ¼ Qmax: But if the quota in the FTA exceeds
Qmax then domestic price will be affected and the FTA is redundant. Again,
whatever welfare is achieved at such a QFTA must have been attainable in the
absence of an FTA and so, by revealed preference, such an FTA cannot be
politically attractive. Thus a necessary condition for an FTA to be politically
attractive is that QFTA, Qmax: This condition can be manipulated as follows:
dg½bða2 PpÞ þ ðgþ bÞða2 PpÞ�
. ðgþ bÞ½bð12 dÞðPp 2 PwÞ þ dðgþ bÞða2 PpÞ2 bdða2 aÞ�
, dgbða2 PpÞ . ðgþ bÞdðgþ b2 gÞða2 PpÞ
þ bð12 dÞðgþ bÞðPp 2 PwÞ2 bdðgþ bÞða2 aÞ
, dgða2 PpÞ . ðgþ bÞ{dða2 PpÞ þ ð12 dÞðPp 2 PwÞ2 dða2 aÞ}
, ðdg2 dðgþ bÞÞða2 PpÞ . ðgþ bÞð12 dÞðPp 2 PwÞ
, dbðPp 2 aÞ . ðgþ bÞð12 dÞðPp 2 PwÞ ð16Þ
This is a necessary condition only, so if Equation 16 does not hold then an FTA
cannot be politically attractive, and this is more likely for low values of d:Turning to the results of the simulations, our earlier intuitions are indeed borne
out. Figure 10.2 shows the difference in welfare in the FTA and without it as d andPp are varied.16 The partner’s price is varied between a; the minimum marginal
cost of the domestic firm, and Pd; the domestic price without an FTA and is
reported as a percentage of that difference. The figure shows that the FTA
becomes less attractive as the partner’s price gets further from the initial price
without the FTA, for most values of d: Further, the FTA is only preferred when dis sufficiently high—in this case above 0.54 or so—and Pp is also high. Also, the
FTA is more likely to be attractive the lower is b; the relative weight the
government places on consumer surplus. This, too, is intuitive. As we suggested
16 The simulations reported here were run using the following parameter values: a ¼ 150; b ¼ 2;
b ¼ 0:1; Pw ¼ a ¼ 25 and g ¼ 0:3: More details are available from the authors on request.
D.H. Feldman and M. Richardson226
earlier, an FTA will only be attractive if it offers a trade-off that is otherwise
unavailable, and what it offers is the ability to increase profits, at any given price,
beyond that which is achievable without an FTA. When b is high the initial
optimal quota is also high, so the domestic price is low. If this price is less than athere is no domestic production at all and an FTA is redundant.
Interestingly, when b is high an increase in d is likely to lead to an expanded
quota. The reason is that when d is low a quota will be imposed to earn quota rents
which are relatively highly valued. This comes largely at the expense of consumer
surplus. As d rises, however, and profits become more valued relative to rents, the
cost of a quota in terms of foregone consumer surplus becomes more acute since
Figure 10.2: Excess of welfare in FTA over initial welfare, with optimal quotas.
Trade Creation and Residual Quota Protection in a Free Trade Area 227
the marginal effect on profits of a tighter quota is smaller than its effect on rents
(given MC . Pw). So, for high b; the attractiveness of a quota to raise profits is
less than its attractiveness to raise quota rents, in a sense, and as profits become
more valued (d rises) so the government finds it more attractive simply to
liberalize the quota entirely.
Our simulations also show that the quota may be either liberalized or tightened
when the FTA is formed. Figure 10.3 illustrates the increase in imports in the FTA
compared to no FTA, when the quota is set optimally in each case. It suggests that
the quota is more likely to be tightened when d is low. Note that for high d and Pp
the policy-maker chooses a prohibitive quota in this setting both within an FTA
and in its absence, hence the difference in imports shown in Figure 10.3 is zero in
Figure 10.3: Excess of optimal quota in FTA over initial optimal quota.
D.H. Feldman and M. Richardson228
those cases. Figure 10.3 also suggests that, for given d; the quota is more likely to
be tightened when the partner’s price is high. This contradicts with Richardson’s
(1993) tariff-protected competitive industry where joining an FTA leads to an
incentive to liberalize tariffs against non-members.
Lastly, the new political equilibrium will never lead to trade creation in this
industry, so long as quota rents are valued at all. The only case in which it might
appear to is where the residual demand curve (D0 in Figure 10.1) in the post-FTA
political equilibrium intersects the marginal cost curve above the partner’s price.
In that case, the domestic firm would produce to where MC ¼ Pp and this output
plus the quota would be less than demand at the price Pp: However, such a
situation could not reflect an optimal quota as the quota could be expanded with no
consequence for domestic price (and hence consumer surplus) or for domestic
profits. Such an expansion, however, would increase quota rents.
10.4. CONCLUSIONS
This chapter has evaluated the positive and political-economy consequences of
FTA formation for a small domestic economy with quota-protected single-firm
industries. In contrast to the competitive case, FTA formation will not create trade
in the single-firm case when the quota is chosen optimally (and may not create
trade even for an arbitrary quota). Nevertheless, an FTA must prove welfare
enhancing in the traditional sense of increasing real national income. Although the
FTA may worsen domestic overproduction relative to the first best optimum (free
trade) the reduction in consumption deadweight loss unambiguously dominates
any extra production loss.
Conventional welfare gains rarely motivate discriminatory liberalization. We
use a political-support function in which the firm’s profits and quota rents are
substitutable to the policymaker tomake three points. First, for an FTA to dominate
the initial quota-distorted equilibrium, the weight on profit must exceed the weight
on revenue in the policymaker welfare function. Second, politically feasible FTAs
require a partner price close to the initial domestic price. Partner prices closer to that
of the low cost producer cost too much political support from foregone profits and
revenues to justify the lower consumer price. Lastly, moving into an FTA may
create an incentive for the policymaker to liberalize the residual quota on non-
members. Thus, FTA formation need not reduce the volume of world trade outside
of discriminatory regional arrangements. With a quota-protected domestic
monopoly, regional arrangements are neither poison nor panacea.
Trade Creation and Residual Quota Protection in a Free Trade Area 229
APPENDIX A
In this section we provide the derivations of Equations 14 and 15 in the chapter:
first, Equation 14 which gives the optimal in-FTA quota in the linear case when
domestic output is positive. The government sets the quota to maximize the
following:
W ¼ bCSþ ð12 bÞr ðA1Þ
where the domestic price is fixed at Pd ¼ Pp thus demand and hence CS is
unchanged as the quota varies. So
dW ¼ ð12 bÞdr ðA2Þ
From the chapter we have
dr
dQFTA¼ d
dp
dQFTAþ ð12 dÞðPp 2 PwÞ
¼ ð12 dÞðPp 2 PwÞ2 dðPp 2 c0Þ ð12Þ
So the optimal quota solves
dr
dQFTA¼ 0 ) c
0¼
1
d{ð12 dÞPw þ ð2d2 1ÞPp} ðA3Þ
Now, in this linear case we have
c0¼ aþ gQ ¼ aþ gðQ2 Q
FTAÞ ) Q
FTA¼ Qþ
1
gða2 c
0Þ ðA4Þ
whereQ is total domestic demand. Substituting Equation A3 into Equation A4 and
simplifying gives
QFTA
¼ QðPpÞ21
dg½ð12 dÞPw 2 da2 ð12 2dÞPp� ðA5Þ
Finally, substituting in forQðPpÞ ¼ ða2 PpÞ=b and simplifying gives Equation 14:
QFTA
¼1
bdg½dðgþ bÞða2 PpÞ þ bð12 dÞðPp 2 PwÞ2 bdða2 aÞ� ð14Þ
What is the maximum value of this FTA quota such that the domestic price is
exactly Pp? To solve this we need to calculate the maximum quota that yields this
D.H. Feldman and M. Richardson230
price when Pd is not fixed. From Equation 6 of the chapter—the firm’s FOC—we
have
ðQ2 QÞP0d þ Pd 2 c
0¼ 0 ) bQþ a2 2bQ2 ðaþ gðQ2 QÞÞ ¼ 0
) Q ¼ða2 a2 bQÞ
{gþ 2b} ðA6Þ
Substituting this into the inverse demand curve and solving gives the domestic price
as a function of the quota:
p ¼1
{gþ 2b}ððgþ bÞða2 bQÞ þ abÞ ðA7Þ
which can be inverted to find the quota that yields a particular domestic price:
Q ¼1
bðgþ bÞððgþ bÞaþ ab2 ðgþ 2bÞpÞ ðA8Þ
So for p ¼ Pp we have Equation 15 of the chapter:
Qmax
¼1
bðgþ bÞ½ðgþ bÞða2 PpÞ þ bða2 PpÞ� ð15Þ
REFERENCES
Cassing, J. H. and Hillman, A. L. (1985). Political influence motives and the choice
between tariffs and quotas. Journal of International Economics, 19, 279–290.
De Melo, J. and Roland-Holst, D. (1991). “Industrial organization and trade liberalization:
evidence from Korea,” in Empirical Studies of Commercial Policy, R. Baldwin, (ed.),
Chicago: University of Chicago Press, Chapter 10.
Dutz, M. A. (1996). “Oligopolistic firms’ adjustment to quota liberalization,” in Industrial
Evolution in Developing Economies, M. J. Roberts and J. R. Tybout (eds.), Oxford:
Oxford University Press, Chapter 6.
Feldman, D. H. (1993). Redundant tariffs as rational endogenous protection. Economic
Inquiry, 31, 436–447.
Feldman, D. H. (1993). The choice between tariffs and quotas as the means of protection
under autocracy. North American Review of Economics and Finance, 4, 165–178.
Trade Creation and Residual Quota Protection in a Free Trade Area 231
Goodman, D. J. and Kaempfer, W. H. (1993). Free trade area formation with quota
protection. North American Review of Economics and Finance, 4, 157–164.
Helpman, E. (1987). Macroeconomic Effects of Price Controls: The Role of the Market
Structure. NBER Working Paper No. 2434.
Hillman, A. L., Tower, E. and Fishelson, G. (1980). On water in the quota. Canadian
Journal of Economics, 13, 310–317.
Kaempfer, W., McClure, H. and Willett, T. D. (1989). Incremental protection and efficient
political choice between tariffs and quotas. Canadian Journal of Economics, 22,
228–236.
Krishna, K. (1989). Trade restrictions as facilitating practices. Journal of International
Economics, 26, 251–270.
Krishna, K. (1993). Theoretical implications of imperfect competition on quota license
prices and auctions. World Bank Economic Review, 7, 113–136.
Krueger, A. O. (1993). Political Economy of Policy Reform in Developing Countries,
Cambridge: The MIT Press.
Lustig, N. (1992).Mexico: The Remaking of an Economy, Washington, DC: The Brookings
Institution.
Meade, J. E. (1955). The Theory of Customs Unions, Amsterdam: North-Holland.
Ng, F. (1996). A Profile of Tariffs, Para-tariffs, Non-tariff Measures, and Economic Growth
in Developing Countries. World Bank, International Trade Division, International
Economics Department.
Ortiz, G. (1991). “Mexico beyond the debt crisis: toward sustainable growth with price
stability,” in Lessons of Economic Stabilization and its Aftermath, M. Bruno, S. Fischer,
E. Helpman, N. Liviatan and L. Meridor (eds.), Cambridge: MIT Press, Chapter 8.
Richardson, M. (1993). Endogenous protection and trade diversion. Journal of
International Economics, 32, 309–324.
Richardson, M. (1995). Tariff revenue competition in a free trade area. European Economic
Review, 39, 1429–1437.
Sweeney, R., Tower, E. andWillett, T. D. (1977). The ranking of alternative tariff and quota
policies in the presence of domestic monopoly. Journal of International Economics, 7,
349–362.
Tybout, J. (2000). Manufacturing firms in developing countries: how well do they do and
why? Journal of Economic Literature, 38, 11–44.
Vousden, N. (1990). The Economics of Trade Protection, Cambridge: Cambridge
University Press.
Whitworth, A. (1980). Monopoly and the case for price controls. Journal of Development
Studies, 17, 80–95.
World Bank (1994). Adjustment in Africa: Reforms, Results, and the Road Ahead.
New York: Oxford University Press.
Young, L. and Anderson, J. (1980). The optimal policies for restricting trade under
uncertainty. Review of Economic Studies, 47, 927–932.
Young, L. and Anderson, J. (1982). Risk aversion and optimal trade restrictions. Review of
Economic Studies, 49, 291–305.
D.H. Feldman and M. Richardson232
CHAPTER 11
The Political Economy of Trade
Sanctions Against South Africa:
A Gravity Model Approach
WILLIAM H. KAEMPFER* and MARTIN ROSS
Department of Economics, Box 256, University of Colorado at Boulder,
Boulder, CO 80309-0256, USA
Abstract
South African trade fell dramatically during the mid-1980s with the application of
anti-apartheid sanctions that partially restricted trade. This chapter develops the
hypothesis that the severity of sanctions is inversely related to the cost of imposing
those sanctions—a law of demand for sanctions consistent with public choice
principles. We estimate a gravity model of trade flows as a function of distance
and economic mass for South Africa. The results show that sanctions were more
restrictive where trade was less important to a country due to its relative size or
distance from South Africa.
Keywords: South Africa, trade sanctions, gravity model, political economy, public
choice
JEL classification: F1
*Corresponding author.
E-mail address: [email protected]
11.1. INTRODUCTION: THE LAW OF DEMAND FOR SANCTIONS
International economic sanctions are a policy tool employed by one government,
or set of governments, in an attempt to persuade another sovereign state to change
some objectionable policy. One implication of sanctions policy, however, is that
not only do the sanctions impose economic (and presumably political) costs on the
target country, they can also be costly to the sending country. In particular,
sanctions can be thought of as a policy funded by an implicit tax that falls
disproportionately on a subset of specific interests in the sending country—
namely those entities with trading interests in the target (Kaempfer and
Lowenberg, 1990, 1992).
Interest group models in public choice theory suggest that an interest is able to
attain government policy objectives only to the extent that its willingness and
effectiveness to pressure for some policy exceeds the willingness and
effectiveness of the other, contrary interests to object to and block the policy
initiative (Becker, 1985). To this end, one can infer that as economic sanctions
impose a rising policy cost upon some set of negatively impacted interests, then
those sanctions will not be as severe. That is, there should be a negatively sloped
demand curve for sanctions. In many applications of sanctions, however, the
domestic costs of an embargo are weighed against having no costly sanctions at
all. In other words, the potential imposition of sanctions can be seen as a
dichotomous policy choice in which either a total trade embargo is applied (e.g.,
Iraq and Cuba), or negligible, low-cost sanctions are applied such as cancellation
of cultural or sport exchanges. Along these lines, Bonetti (1997) finds that the
probability of imposing trade sanctions is inversely related to the extent of
presanction trade flows between two countries.
One much different case of the imposition of sanctions was that of sanctions
against South Africa’s apartheid regime in the 1980s. Most western countries as
well as the African neighbors of South Africa adopted some form of sanctions
policy against the white regime in South Africa in an attempt to coerce that
government to end political repression of its majority population groups and
democratize. However, a uniform multilateral trade embargo policy was not
adopted. Rather, in this case the political process within each trade partner
established some sort of national package of partial sanctions, three features of
which are noteworthy. First, the sanctions policies were internally decided upon
by the sanctioning nations, not by some multi-lateral authority such as the UN or
Organization of Commonwealth States. Second, the sanctions were not uniform
across countries. Third, anti-apartheid sanctions were not a dichotomous policy,
but instead a variety of specific goods categories were embargoed in various
countries (Kaempfer and Lowenberg, 1989). As a result, sanctions against
W.H. Kaempfer and M. Ross234
South Africa allow the opportunity to test the proposition that as the cost of
sanctions rises, less severe sanctions will be imposed.
In this chapter, we undertake such a test by means of applying a gravity model
to South African trade before and after the time of the major imposition of
sanctions in 1986. Gravity models (see, for instance, Bergstrand, 1985) explain
trade as a function of mass and distance where greater economic mass (i.e., GDP
and/or per capita GDP) and geographic closeness (i.e., lower transport costs) will
lead to a greater volume of trade between countries. For the purposes of our point,
those same attractors—mass and closeness—raise the political costs of applying
sanctions. Consequently, we expect to find that the trade for which there is the
least “gravitational attraction” will be the most impacted by sanctions, and
consequently we hypothesize that the gravity model will explain South African
trade better in the late 1980s relative to the early 1980s. This hypothesis is also
tested by examining changes in the effects of geographical proximity over time.
11.2. SOUTH AFRICAN TRADE DURING THE 1980S
South Africa’s economy in the 1980s, while largely mineral based, had a number
of unusual features. Decades of apartheid-induced partial isolationism led the
country to have a number of large parastatal economic entities that were not
particularly economically efficient. However, a relatively large upper income
group led to a great deal of imports from other high-income countries, especially
in manufactured goods. At the same time, beyond the mining section, a
manufacturing sector in South Africa served some export markets, and a southern
hemisphere location allowed for “off-season” agricultural exports of fruits and
vegetables. Overall, the volume of trade (imports and exports) amounted to 60%
of GDP in 1981. This relatively large proportion of trade in the economy provided
a vehicle for other nations to alter the behavior of the South African government.
The 1986 anti-apartheid sanctions against South Africa were imposed by
several important trading partners, including the United States, United Kingdom,
European Economic Community, Canada and other Commonwealth countries and
Japan. The sanctions involved areas such as disinvestments, bank loans, tourism
restrictions, and import and export restrictions. The focus here is on the trading
restrictions, the majority of which involved restricting South Africa’s exports to
one or more of its trading partners—specifically of steel, iron, coal, uranium,
textiles, Krugerrands and agricultural products, as well as some limited imports
generally in the areas of arms and technology goods.
Complete data series are only available for three African countries (Kenya,
Malawi and Zimbabwe). These nations are the three largest African trading
Political Economy of Trade Sanctions Against South Africa 235
partners of South Africa and are relatively close geographically to South Africa.
Data for the top 14 non-African trading partners are also examined and show the
trading trends with the rest of the world (ROW). (Note: in the early 1980s, Sweden
had a large trade volume with South Africa. However, as a strong supporter of the
sanctions, their trade after 1986 declined to the point where it is difficult to obtain
the exact values.)
Figure 11.1 shows the aggregate trade, imports and exports of South Africa with
the three African countries and the 14 other countries (ROW). Values are in 1985
South African Rand. Figure 11.2 presents the figures with the individual African
nations, while Figures 11.3–11.6 give the trade amounts for the remaining 14
Figure 11.1:
W.H. Kaempfer and M. Ross236
included nations. There is a large variation in the relative sizes of the illustrated
trade flows—the US is the largest, followed by Germany, Japan and the United
Kingdom.
Table 11.1 summarizes the information on trade which is presented in the
figures. The table shows the percentage changes in trade from 1981 to 1985 and
from 1986 to 1990 (where the denominator used in calculating the percentage
changes in both columns in the table is the value of trade in 1981). These values
show that South Africa’s trade with the world as a whole was declining
significantly from 1981 to 1985.
11.3. THE GRAVITY MODEL OF TRADE
Since their development, gravity models have been empirically successful at
explaining the volume of trade flows between countries. Tinbergen (1962) and
Poyhonen (1963) were the first to propose these econometric models of bilateral
trade. The models explained the volume of trade between two countries
Figure 11.3:
W.H. Kaempfer and M. Ross238
as a function of their GNPs and the distance between them. The GNP variable
represents the economic mass of a country and thus the cumulative income
elasticity impact on trade, while distance is a proxy for transport costs. Linnemann
(1966) included a population variable in the analysis to account for scale
economies. Leamer (1974) used this expanded gravity model to test the effects of
factor endowments and other country characteristics on trade flows. While these
models were quite accurate and possessed an intuitive appeal, they lacked a strong
theoretical framework. Bergstrand (1985, 1989) removed some of these concerns
by extending the basic model to include a microeconomic foundation.
This chapter adopts the novel approach of employing the gravity model as a
method of examining the severity of the sanctions against South Africa. The
hypothesis is that the severity of sanctions should be inversely related to cost. The
gravity model predicts that countries close to South Africa or countries with
higher income (either total or per capita) will have more trade with South Africa.
By similar reasoning, these countries will find it more politically costly to restrict
trade via sanctions. Consequently, following the imposition of differential
Figure 11.4:
Political Economy of Trade Sanctions Against South Africa 239
sanctions, the predictive power of the gravity model will increase as more distant
and lower income countries will find it easier to impose more severe sanctions
policies. We test this idea by examining a traditional gravity model in two separate
time periods for South African trade data—from 1981 to 1985 in the period before
the most publicized sanctions were enacted and from 1986 until 1990.
The estimated equation of gross trade flows between South Africa and an
individual country i is a function of the size of the two economies as well as the per
capita income levels and the distance between countries (following the standard
literature, all variables are in logs to remove heteroskedasticity)1:
Tradei ¼ b0 þ b1 South African GDP þ b2 per capita South African GDP
þ b3 GDPi þ b4 (per capita GDP)i þ b5 distancei þ e i
Figure 11.5:
1 The values for trade come from the International Trade Statistics Yearbook and are in 1985 US
dollars. The GDP, exchange rate and population figures are from the International Financial Statistics
Yearbook, again with dollar amounts expressed in 1985 South African Rand. Nautical distances are
from the US Naval Office and land distances are from Rand McNally.
W.H. Kaempfer and M. Ross240
The effect of sanctions may be captured in two ways. First, the gravity equation
may explain a different proportion of the variation in trade flows across the two
time periods. Second, there may be a significant difference between the parameter
estimates of the distance and foreign income variable before and after sanctions.
11.4. EMPIRICAL RESULTS
Table 11.2 presents the regression results, where the parameter estimates indicate
elasticities. The gravity equation does a fairly good job of predicting the trade
Figure 11.6:
Political Economy of Trade Sanctions Against South Africa 241
flows between South Africa and the ROW. While the GDP variables for the
trading partners of South Africa are significant, South African GDP and per capita
GDP are not. This may be due to the relatively large amount of trade in the South
African economy and the small variation in these variables over the sample.
Table 11.1: Percentage changes in trade with South Africa.
1981–1985 1986–1990 1981–1990
Africa 217 24 220
ROW 24 217 220
Kenya 158 278 80
Malawi 58 228 30
Zimbabwe 230 1 230
Australia 28 272 244
Austria 211 64 53
Belgium 13 19 32
Canada 230 220 250
France 224 25 229
Germany 0 5 6
Israel 75 212 63
Italy 6 23 3
Japan 26 225 231
The Netherlands 93 2127 234
Spain 49 29 78
Switzerland 228 224 251
United Kingdom 29 27 215
United States 28 231 240
Table 11.2: Regression results (dependent variable is trade with country i).
Variable 1981–1985 1986–1990
Constant 3.96 (0.06) 17.23 (0.62)
South African GDP 20.18 (20.01) 21.38 (20.28)
Per capita South African GDP 0.17 (0.01) 0.42 (0.11)
GDPi 0.87* (7.56) 0.60* (6.56)
(Per capita GDP)i 0.03 (0.17) 0.54* (3.69)
Distancei 20.89* (24.70) 21.30* (27.57)
F-statistic 22.16 39.25
R2 0.58 0.71
Adjusted R2 0.56 0.70
t-statistics in parentheses. *significance at the 1% level.
W.H. Kaempfer and M. Ross242
The gravity model states that a country’s trade with a close, large, high-income
country will be greater than trade flows to a distant, small, low-income country.
However, the application of sanctions by a close, large, high-income country will
be more costly to that country due to the greater amount of trade that country had
initially with its trading partner. Many factors explain the volume of trade between
two nations. However, after the sanctions are enacted, the estimated importance of
economic mass and proximity may increase since sanctions may tend to be less
restrictive where trade is important. These political costs of applying sanctions
should lead to an increase in the predictive power of the gravity model after
sanctions take effect.
The adjusted R2 is 0.56 for the time period 1981–1985 and 0.70 for 1986–
1990. A comparison of these two values indicates that the gravity model explains
more of the variation in trade after the sanctions takes effect. This suggests that the
sanctions were more restrictive where trade was less important and implies the law
of demand for sanctions.
Another indication of the downward sloping demand for sanctions comes
from comparing the parameter estimates on the distance and foreign GDP
variables in the two estimated equations. After 1986, if the predictive power of
the gravity model is improved, then the influence of distance and GDP on trade
should increase since the most restrictive sanctions should occur where
restricted trade with South Africa is least costly. The results in Table 11.2 show
that the effect of the distance variable is more negative from 1986 to 1990 than
it is from 1981 to 1985. A t-test shows that there is a significant difference
between the two estimates. The results show that the increasing distance has a
significantly more negative effect on trade after the sanctions took effect.
Similarly, when taken together, the two foreign income variables show a greater
positive effect on trade after the sanctions. The coefficient on foreign GDP
declines slightly, though not significantly, between the two estimated equations,
while the foreign per capita GDP variable only becomes significantly positive
in the second time period.
11.5. CONCLUDING REMARKS
The use of trade sanctions imposes costs not only upon the target of those
sanctions but also on the sanctioning countries as well. Ceteris paribus, the costs
of imposing sanctions will rise as the trade linkages between sender and target
rise. Furthermore, these costs are likely to have a political manifestation that will
make severe sanctions more difficult to impose when sanctions costs are high on
the sending nation.
Political Economy of Trade Sanctions Against South Africa 243
In this chapter, we examined South African trade during the 1980s, a period of
great sanctions turbulence, by means of a gravity model. Our results support the
hypothesis that the post sanctions trade interruptions increase the explanatory
power of the gravity model. This suggests, in accordance with public choice
principles, that sanctions against South Africa were most severe where
interruptions of trade with South Africa were least costly.
ACKNOWLEDGEMENTS
We are grateful to Ed Tower and Ed Balistreri for their help. Tony Lowenberg’s
advice and guidance was invaluable.
REFERENCES
Becker, G. (1985). Public policies, pressure groups, and dead weight costs. Journal of
Public Economics, 28, 329–347.
Bergstrand, J. (1985). The gravity equation in international trade: some microeconomic
foundations and empirical evidence. Review of Economics and Statistics, 67,
474–481.
Bergstrand, J. (1989). The generalized gravity equation, monopolistic competition, and the
factor-proportions theory in international trade. Review of Economics and Statistics, 71,
143–153.
Bonetti, S. (1997). A test of the public choice theory of economic sanctions. Applied
Economics Letters, 4, 729–732.
International Financial Statistics Yearbook, International Monetary Fund, various years.
International Trade Statistics Yearbook, International Monetary Fund, various years.
Kaempfer, W. and Lowenberg, A. (1989). Sanctioning South Africa: the politics behind the
policies. Cato Journal, 8, 713–727.
Kaempfer, W. and Lowenberg, A. (1990). “Analyzing economic sanctions: toward
a public choice framework,” in International Trade Policies: Gains from Exchange
between Economics and Political Science, Studies in International Trade Policies, T.
Willett and J. Odell (eds.), Ann Arbor, MI: University of Michigan Press, 173–193.
Kaempfer, W. and Lowenberg, A. (1992). International Economic Sanctions: A Public
Choice Perspective, Boulder, CO: Westview Press.
Leamer, E. (1974). The commodity composition of international trade in manufactures: an
empirical analysis. Oxford Economic Papers, 350–374.
Linnemann, H. (1966). An Economic Study of International Trade Flows, Amsterdam:
North-Holland.
W.H. Kaempfer and M. Ross244
Poyhonen, P. (1963). A tentative model for the volume of trade between countries.
Welwirtschaftliches Archiv, 90, 93–99.
Rand McNally Atlas, Rand McNally and Co.
Tinbergen, J. (1962). Shaping the World Economy: Suggestions for an International
Economic Policy, New York: McGraw-Hill.
US Naval Oceanographic Office. Distance Between Ports, H.O. Publication No. 151. US
Government Printing Office.
Political Economy of Trade Sanctions Against South Africa 245
CHAPTER 12
The Political Economy of
Unconditional and Conditional Foreign
Assistance: Grants vs. Loan Rollovers
WOLFGANG MAYERa,* and ALEX MOURMOURASb
aDepartment of Economics, University of Cincinnati, Cincinnati, OH 45221-0371, USAbIMF Institute, International Monetary Fund, Washington, DC 20431, USA
Abstract
Improving the effectiveness of assistance programs is a priority of international
financial institutions (IFIs). This chapter examines the effectiveness of alternative
assistance instruments in a dynamic political-economy framework. Economic
policies of the receiving country are distorted by the influence of a domestic
interest group. The assistance-providing IFI aims at reducing these distortions.
The IFI provides assistance either as grants or loans, and either conditionally on
reducing policy distortions or unconditionally. The chapter shows that, other
things constant, one-time grants are more effective than loan rollovers when
assistance is unconditional, but that this effectiveness ranking is reversed when
assistance is conditional.
Keywords: International Monetary Fund, conditionality, common agency models,
grants vs. loans
JEL classifications: E61, F33, F34
*Corresponding author.
E-mail address: [email protected]
12.1. INTRODUCTION
Developing countries have received large amounts of international financial
assistance over the last 50 years. Much of this assistance has been channeled from
developed to developing countries under the auspices of International Financial
Institutions (IFIs). In some instances, resource transfers have had lasting benefits
to the recipient countries. In many other cases, however, enduring success has
eluded successive IFI-supported economic reform programs. Hence, it comes as
no surprise that IFIs have considered redesigning their assistance programs to
achieve better results. A general goal of this chapter is to gain new insights about
the effectiveness of different forms of assistance.
Two important policy concerns in the design of IFI-supported programs are the
length of IFI financial involvement in recipient countries and its “payback
provisions.” Specifically, at issue is whether assistance ought to continue in the
form of (often concessional) loans or whether it ought to be converted to grants
that do not have to be repaid. The length of IFI financial involvement has a direct
bearing on the issue of inappropriate “prolonged use” of IFI resources. Prolonged
users are transition or developing countries whose economic reform and
adjustment programs have been supported extensively by the IFIs through
successions of conditional loans that have often spanned two or more decades.1
While prolonged financial association with IFIs has in some cases been beneficial
to low-income developing and transition countries, in other cases the succession
of low-interest IFI loans has failed to yield lasting improvements in policies and
economic performance.2 Better ex post monitoring of reform programs with
identified prolonged users and graduation or “exit” from financial engagement in
cases of inappropriate prolonged use is now an explicit IFI goal.3
The repayment terms of assistance are a second major policy issue in the
design of both bilateral and multilateral assistance programs. Several decades
ago, Friedman (1958, p. 515) proposed to substantially increase US
development assistance and turn it into a “final terminal grant…[that] should
be something like two to three times the annual grants we have been making to
the country.” This issue of grants vs. loans resurfaced again recently after US
President George W. Bush proposed to turn into grants one half of the
assistance provided through the International Development Association (IDA),
1 See IMF (2002) for a detailed definition of prolonged use and for the list of countries identified as
prolonged users of IMF resources.2 Bird et al. (2000) and IMF (2002, 2003a) contain detailed discussions of the drawbacks of prolonged
use. IMF (2003b) offers guidance on how to avoid inappropriate prolonged use of IMF resources.3 See, for instance, IMF (2002, p. 328).
W. Mayer and A. Mourmouras250
the arm of the World Bank that provides assistance to qualified low-income
countries. While the World Bank welcomed the proposal, it was met with fierce
resistance elsewhere. Some bilateral donors expressed concerns that, in the
absence of additional funds to replenish its capital, a switch to grants would
deplete IDA capital and compromise its ability to provide assistance to poor
countries in the future. Eventually the G-7 agreed in a June 2002 meeting of
their finance ministers in Halifax to increase the proportion of IDA funding that
is provided in the form of grants to between 18–21%, as Cunningham (2002)
points out.
This chapter assesses the merits of different forms of IFI financial involvement,
including their length and repayment terms (whether grants or loans), with the
help of a dynamic version of the political-economy model developed by Mayer
and Mourmouras (2002). They apply the common agency framework of
Grossman and Helpman (1994) and Dixit et al. (1997) to study the effects of
conditional IFI assistance programs in the presence of domestic interest groups.
Interest groups resist reforms when their privileges (rents) are impaired by the
implementation of efficiency-enhancing policy measures. The IFI uses its
financial leverage to steer member governments toward policies that promote
domestic and international prosperity, an assumption that is consistent with the
IFIs’ stated purposes.4 For assistance programs to succeed, IFIs must fully account
for the influence of domestic interest groups on the policy choices of assistance-
receiving governments. IFI assistance helps shift the political equilibrium in
recipient countries and improve the quality of policies selected, thus providing
a countervailing influence to the power of domestic special interests.5
In this framework, we compare a one-time “final” grant with a succession of
one-period IFI loans of the same size that, for simplicity, are assumed to be
interest-free. Whereas the grant cannot be recalled once its conditionality is met
and the IFI has released its funds, assistance that takes the form of loans can be cut
off (not be renewed) if the recipient government does not adhere to the agreed
conditions. The effectiveness of assistance is judged from the IFI’s point of view.
The IFI acts on behalf of the world community, with the objective of maximizing
world welfare. It can raise world welfare in two ways: one is to transfer resources
from the rest of the world (ROW) to the developing country; the other is to create
4 See, for instance, the IMF’s Articles of Agreement, especially Article I, Section 5.5 The experience with reforms in some countries of the Commonwealth of Independent States in the
1990s is a particularly good example of the resistance of special interests to reforms. During the early
years of transition, “red directors” and other special interests used their political influence to evade
taxes, obtain subsidies, strip the assets on state enterprises and extract other special privileges. The IFIs
provided assistance to pro-reform governments while being aware of the need of reformers to stay in
power.
Unconditional and Conditional Foreign Assistance 251
incentives for the assistance-receiving country to reduce its policy-created
distortions. The developing country’s incumbent government chooses economic
policies under the influence of a domestic interest group. Its objective is to
maximize political support that comes from both the interest group and the general
public.
An advantage of assistance through loan rollovers is that the IFI can adjust
its assistance package as economic and political circumstances change over
time. The amount of a loan depends on its benefit in the receiving country and
its cost to the ROW, as well as on the recipient government’s concern for its
general public. If, at time of repayment, any of these influences has changed,
the IFI can replace the repaid loan with a new one that differs in size. Should all
economic and political parameters remain unchanged, the IFI’s interests are best
served by simply reissuing the repaid loan. A grant, on the other hand, cannot
be adjusted ex post. Once made, all transferred resources are permanently
shifted from the ROW to the developing country. No matter how the IFI’s
incentives to support a developing country might have changed, the grant
cannot be recalled.
An advantage of assistance through a grant is that the set of projects for which
transferred resources can be employed is larger than for loans. The need to repay
a loan after a specified time interval limits loan-financed investments to short-
term projects. Knowing that assistance through grants cannot be reversed, on the
other hand, enables the receiving country to invest in both short- and long-term
projects.
The specific purpose of this chapter is to highlight the implications of a third
difference between loan rollovers and a one-time “final” grant, namely, of their
“commitment value”. The interactions between an aid-providing IFI, an aid-
receiving government, and an interest group unfold in a multi-period game.
Assistance through loans can be periodically revised; assistance through grants
cannot. Accordingly, the grant represents a stronger commitment on the part of the
IFI. At issue is whether this difference in commitment value introduces a bias in
favor of either the policy of loan rollovers or a one-time grant, judged from the
perspective of the assistance-giving IFI. Stated differently, is world welfare higher
when the IFI provides assistance through loan rollovers or through a one-time
grant, assuming away all other advantages and disadvantages of the two
instruments?
Our answer to the above question is clear-cut: the IFI is better off assisting
through a final grant than through loan rollovers when assistance is unconditional;
but the IFI is better off assisting through loan rollovers than a final grant when
assistance is conditional. These conclusions are derived from a model in which the
IFI compares the present value of world welfare resulting from a grant with the
W. Mayer and A. Mourmouras252
present value of welfare resulting from a succession of loans.6 The model captures
the influence of an interest group on the developing country’s government by
adopting the Dixit et al. (1997) formulation of a truthful equilibrium. The interest-
group-influenced government plays a game with the IFI both when assistance is
unconditional and when it is provided conditionally. When assistance is
unconditional and takes the form of loans that are being rolled over, the
government chooses its economics policies and the IFI chooses its loan at the
same time, namely, at the beginning of each period. In the case of a “final” grant,
the government still chooses its policies at the beginning of each period, but the
IFI can choose the grant value only once, namely, at the beginning of the initial
period.
When assistance is conditional, the nature of the political game changes
dramatically as the IFI joins the interest group as a principal in a common-agency
game. The implications for welfare of the IFI are drawn from the conditions of a
truthful equilibrium. In the case of loan rollovers, government, interest group, and
IFI play a two-stage game at the beginning of each period. The IFI is thus engaged
in a prolonged financial association with the recipient country. In the case of a
grant, on the other hand, the three players engage in a two-stage game only at the
beginning of the initial period. While the IFI accounts for government responses
in succeeding periods, it can act only during the initial period. For all periods
beyond the initial one, only the government and the interest group can take
actions. By definition, the policy of providing a final grant does not lead to a
prolonged financial association between the IFI and the recipient country.
The conclusion that a final grant is superior, from the IFI’s point of view, to an
indefinite rollover of a loan of the same amount is not surprising for the case of
unconditional assistance. The higher commitment value of the one-time grant
conveys a benefit to the IFI. As it chooses the grant level prior to the government’s
choices of policies beyond the initial period, the IFI has some first-mover
advantages in the unconditional assistance game. Under conditional assistance,
loan rollovers are superior to a grant because it enables the IFI to enforce
conditionality in each period. A critical feature of the political equilibrium under
conditional assistance is that it results in Pareto-optimality. The outcome is a
combination of assistance and interest group-influenced economic policies that
maximizes joint welfare of the three players, namely, government, interest group,
and IFI. When assistance takes the form of one-period renewable loans, the IFI’s
ability to enforce its conditions on economic policies assures Pareto-optimality.
6 An alternative political economy model is formulated by Drazen (2002) whose government contends
with veto players; i.e., with constitutional actors that influence policy-making from within the
government.
Unconditional and Conditional Foreign Assistance 253
When assistance takes the form of a one-time grant, on the other hand, the
governmentmust adopt distortion-reducing policies only during the initial period in
order to receive the grant. The IFI’s ability to enforce economic policy conditions
disappears after the initial period, as the grant results in an irreversible transfer of
resources.After the initial period, the grant-receiving government takes the value of
the grant as given and views it the same way as an unconditional grant of the same
value. The choices of assistance by the IFI and economic policies by the
government are then no longer Pareto-optimal. Furthermore, this departure from
Pareto-optimality under a conditional grant, when contrasted to the Pareto-optimal
outcome under conditional loan rollovers, comes at the cost of reduced IFI welfare.
The IFI is worse off under a conditional grant than under conditional loan rollovers.
12.2. THE COMMON AGENCY MODEL
12.2.1. Decision makers and their objectives
Consider a developing country whose economic policies are shaped by the
interactions of three different decision makers. First, an incumbent domestic
government chooses the country’s economic policies. Second, a domestic interest
group that benefits from policies that distort the economy attempts to influence the
policy maker. Third, an IFI that acts as gatekeeper of the world’s welfare provides
economic assistance. This assistance benefits the developing country directly as it
increases its capital stock, as well as indirectly as it lowers the level of economic-
policy-generated distortions.
Welfare of the developing country is measured by its national income. Given
the country’s endowment with labor and capital, national income depends on the
amount of economic assistance received, A; and the degree of economic
distortions generated by economic policies. The economic assistance enables the
country to expand its capital stock. The degree of distortions is measured by an
index 0 # v , 1; and it is assumed that the distortion-caused loss in national
income rises linearly with the distortion index. The developing country’s welfare,
W ; therefore, is
W ¼ ð12 vÞyðAÞ ð1Þ
where yð0Þ measures potential national income in the absence of distortions and
with no assistance, and where y0ðAÞ . 0 and y00ðAÞ , 0 for A $ 0:The incumbent government chooses the economic policies of the developing
country. Its objective is to adopt policies, and a corresponding distortion index v;
W. Mayer and A. Mourmouras254
to maximize its political support from the country’s interest group and general
public. As in Grossman and Helpman (1994), the government receives political
support from the interest group in the form of financial contributions, C; and fromthe general public in form of “approval”. The latter depends on the country’s
overall welfare, W : In making policy choices, the government faces conflicting
attitudes of interest group and general public; the interest group benefits from
policies that are more distorting—such as tariffs, quotas, monopolies, subsidies,
etc.—while the general public is hurt by distortions. To pursue its goals, the
interest group tenders a financial contribution schedule, CðvÞ; to the government.
This schedule makes the amount of financial support contingent on the
government’s choice of distorting policies that favor the interest group. The
government’s political support function, therefore, is written as
G ¼ CðvÞ þ að12 vÞyðAÞ ð2Þ
where a $ 0 is a parameter that reflects the government’s concern for welfare of
the general public. Its value depends on the government’s dependence on the
goodwill of the public. When the developing country’s political institutions are
weak, the value of a tends to be small.
There is only one interest group in the developing country. It benefits from
certain policies that are distorting and, therefore, pressures the government to adopt
them. The interest group’s net welfare,V; equals utility obtained from the distorting
policies, UðvÞ; minus its financial contribution to the government, CðvÞ; i.e.,
V ¼ UðvÞ2 CðvÞ ð3Þ
where the group’s welfare without contributing is assumed to rise with the
distortion index at a decreasing rate. Hence, U 0ðvÞ . 0 and U 00ðvÞ , 0: We also
assume that limv!0 U0ðvÞ ¼ 1 and limv!1 U
0ðvÞ ¼ 0: These assumptions assure
that the economy is always riddled with distorting policy choices as along as there
is an interest group.
Finally, there exists an international financial institution (IFI). It is an
institution that was set up in the past by the entire world community. Its intended
mission is to serve as a public-interest institution and to maximize world welfare
by channeling assistance from the ROW to a developing country. Welfare of the
ROW is also measured by its national income. In contrast to the developing
country, however, distorting economic policies have a negligible impact on the
ROW’s total output. Given the ROW’s factor endowments, its welfare,Wp; solelydepends on the amount of assistance provided, such that
Wp¼ y
pð2AÞ ð4Þ
Unconditional and Conditional Foreign Assistance 255
where ypð0Þ is the ROW’s output in the absence of assistance, yp0ð2AÞ . 0 and
yp00ð2AÞ , 0 for all A $ 0: Given the welfare measures of the developing country
and ROW, the IFI’s objective function is
I ¼ ð12 vÞyðAÞ þ ypð2AÞ ð5Þ
12.2.2. Political equilibrium with unconditional assistance
The nature of the developing country’s political equilibrium critically depends on
whether the IFI provides assistance conditionally or unconditionally. Assistance is
conditional if the IFI makes the amount of aid contingent on the adoption of
distortion-reducing economic policies. As explained by Mayer and Mourmouras
(2002), the conditional assistance scenario can be conveniently modeled within
the political-economy, common-agency framework of Grossman and Helpman
(1994) and Dixit et al. (1997). The government acts as the common agent of
domestic interest group and IFI. The latter present the government with a
contribution and assistance schedule, respectively, to press for their opposing
interests in economic policies. Assistance is unconditional if its provision by the
IFI is not contingent on the government’s adoption of distortion-reducing policies.
Although the IFI does not impose conditions, it is fully aware that world welfare
depends on the assistance-receiving country’s policy distortions. The IFI,
therefore, reacts to the government’s policy choices even though there is no
contractual agreement between IFI and government. Accordingly, the political
game itself involves only the developing country’s government and its interest
group when assistance is unconditional.
Under unconditional assistance, the government’s choice of policies is the
outcome of a two-stage game in which the government chooses the distortion
index (economic policy) in the second stage given the financial contribution
schedule tendered by the interest group in the first stage. We are focusing on
equilibria that are truthful; i.e., on equilibria for which the contribution schedule of
the interest group is truthful relative to the equilibrium welfare level of the players.
The general conditions for a truthful political equilibrium are stated in Proposition
3 of Dixit et al. (1997). Adapting these conditions to our situation of a political
game in which only one interest group and the government participate, the truthful
political equilibrium consists of a truthful financial contribution schedule, CT; anda distortion index, v0; that is characterized by two conditions: first, the
government’s choice of v0 must be such that v0 ¼ argmax0#v,1 ½CTðv;V0Þ þ
að12 vÞyðAÞ�; where V0 denotes the interest group’s net welfare in equilibrium
and A is the amount of unconditional aid received when there is an interest group.
W. Mayer and A. Mourmouras256
Second, ½CTðw0;V0Þ þ að12 v0ÞyðAÞ� ¼ að12 v2V ÞyðA2V Þ; where v2V is the
distortion index the government would choose in the absence of influence seeking
by the interest group and A2V is the amount of unconditional aid received in the
absence of an interest group. This second condition states that the interest group’s
truthful contribution is just enough to make political support for the government
the same with influence seeking as it would be without influence seeking. With no
interest group—as expressed by the right-hand side of the equation where C is set
equal to zero—the government would choose distortion-free economic policies
such that v2V ¼ 0 and the IFI would provide A2V of aid to this distortion-free
economy. One also can see from Equation 3 that CTðv;V0Þ ¼ UðvÞ2 V0:Consequently, the first of above conditions requires that v0 is chosen to satisfy
U0ðv0
Þ2 ayðAÞ ¼ 0 ð6Þ
whereas the second condition implies that the interest group’s equilibrium
contribution equals
CTðv0;V0
Þ ¼ a½v0yðAÞ þ yðA
2VÞ2 yðAÞ� ð7Þ
Clearly, the degree of distortions, v0 depends on both the amount of unconditional
assistance received from the IFI, A; and the government’s concern for the general
public’s welfare, a: As can be seen from Equation 6, the more unconditional aid is
received and the more the government cares for the public’s welfare, the less-
distorting policies are adopted. This response of the government to different
assistance levels is traced out in Figure 12.1 as the RGRG curve. Since U 00ð·Þ , 0;the government’s policy reaction curve is downward sloping.
12.2.3. Political equilibrium with conditional assistance
Assistance is conditional when the IFI makes its aid to the developing country
contingent on the government’s adoption of less-distorting policies. The IFI,
thereby, becomes a second principal in the economic-policy game. As before, the
government chooses economic policies and the corresponding distortion index.
But unlike the unconditional assistance scenario, the interest group’s pressure for
more-distorting policies is now counteracted by the IFI that, in the interest of
world welfare, pushes for less-distorting policies. We now add an assistance
schedule tendered by the IFI to the contribution schedule of the interest group.
Although both payment schedules are offered to the government, an important
difference between them is that interest group contributions raise the govern-
ment’s political support directly whereas assistance payments benefit the
Unconditional and Conditional Foreign Assistance 257
government only indirectly. The interest group contribution goes into the
campaign funds or personal pockets of politicians that constitute the government.
The assistance payment, on the other hand, goes in its entirety to expand the
economy’s production potential. The raised production potential enlarges national
income that, in turn, leads to stronger approval of the government by the general
public.
The conditional assistance model again adopts the common-agency approach
first developed by Bernheim and Whinston (1986) and later applied and further
refined by Grossman and Helpman (1994) and Dixit et al. (1997). The government
is viewed as the common agent of interest group and IFI. They play a two-stage
economic-policy game in which the government chooses policies in the second
stage given the contribution schedule of the interest group and the assistance
schedule of the IFI. Both payment schedules are offered to the government in the
first stage. Again we are in search of an equilibrium in which both contribution
schedule, CT; and assistance schedule, AT; are truthful. And again we employ the
conditions of Proposition 3 of Dixit et al. (1997) to characterize this truthful
equilibrium. The first condition now is that the government’s policy choice, v1; issuch that v1 ¼ argmax0#v,1 {C
Tðv;V1Þ þ að12 vÞy½ATðv; I1Þ�}; where V1 and
I1; respectively, denote the interest group’s net welfare and the entire world’s
(IFI’s) welfare evaluated at the conditional assistance equilibrium. Second, the
Figure 12.1: Conditional assistance equilibrium.
W. Mayer and A. Mourmouras258
truthful contribution schedule of the interest group in equilibrium must satisfy
{CTðv1;V1Þ þ að12 v1Þy½ATðv1; I1Þ�} ¼ að12 v2V Þy½ATðv2V ; I1Þ�; where
v2V is again the government’s choice of the distortion index when the interest
group does not contribute. As was the case in the unconditional assistance model,
v2V ¼ 0: Third, the truthful assistance schedule of the IFI must satisfy
{C T(v 1,V 1) þ a(1 2 v 1)y[A T(v 1, I 1)]} ¼ {C T(v 2I,V 1) þ a(1 2 v 2I)y(0)},
where v2I. 0 would be the government’s choice of the distortion index if the IFI
did not offer any conditional assistance. The first condition states that the
government chooses a policy that, given the truthful contribution schedule of the
interest group and the truthful assistance schedule of the IFI, maximizes its
political support. The second and third conditions spell out how much interest
group and IFI, respectively, contribute. The interest group’s truthful equilibrium
contribution must be such that political support for the government is as strong
when it contributes as it would be if it did not contribute, whereby the
government-adopted policies entail distortion index v2V ¼ 0 and the IFI is just as
well off as in equilibrium. The IFI’s truthful equilibrium assistance payment must
be such that political support for the government is as strong when the IFI assists
as it would be if it made no conditional assistance payment, whereby the
government chooses distortion index v2I ; and the interest group is just as well offas in equilibrium.
Recalling again that CTðv;V1Þ ¼ UðvÞ2 V1; the first condition for a truthful
equilibrium requires that v1 is chosen such that
2U 0ðv1Þ2 ay½ATðv1; I1Þ�
að12 v1Þy0½ATðv1; I1Þ�¼
2y½ATðv1; I1Þ�
yp 0½2ATðv1; I1Þ�2 ð12 v1Þy0½ATðv1; I1Þ�ð8Þ
The left-hand side of Equation 8 reflects the slope of the government’s political-
support indifference curve, ›Að·Þ=›v; derived from Equation 2 after substitution
of CðvÞ ¼ UðvÞ2 V : It states the rate at which the government is willing to
accept more economic assistance for fewer distortions. The right-hand side of
Equation 8, on the other hand, expresses the slope of the IFI’s world-welfare
indifference curve, ›Að·Þ=›v; derived from Equation 5. It states the rate at which
the IFI is willing to offer more assistance for reduced policy distortions.
Consequently, the political-support-maximizing choice of the distortion index,
v1; implies that joint welfare of domestic government and IFI are maximized.
The second condition for a truthful equilibrium requires that the interest group’s
financial contribution is such that
Uðv1Þ2 V
1þ að12 v1
Þy½ATðv1; I1Þ� ¼ ay½AT
ð0; I1Þ� ð9Þ
Unconditional and Conditional Foreign Assistance 259
where we substituted for v2V ¼ 0: Correspondingly, the third truthful equilibriumcondition requires that the IFI’s assistance payment is such that
Uðv1Þ þ að12 v1
Þy½ATðv1; I1Þ� ¼ Uðv2I
Þ þ að12 v2IÞyð0Þ ð10Þ
where v2I is the government’s choice of distortions when A ¼ 0:Figure 12.1 portrays the equilibrium choice of economic assistance, ATðv1; I1Þ;
and economic distortions, v1; when assistance is conditional. The diagram
highlights the interactions between IFI and government while keeping the role of
the interest group in the background. It is implicitly assumed that the IFI has no
incentive to offer any unconditional assistance; if there is any assistance at all, it is
conditional. Concerning the diagram, we first note the already-mentioned RGRG
locus. It is the government’s best-response function to the IFI’s provision of
assistance. In the absence of IFI assistance, the government pays attention to the
wishes of the domestic interest group only and chooses distortion index v2I : TheGG curve traces out those combinations of distortion index and IFI assistance that
yield a constant level of political support, given the interest group’s contribution
function. The reflected level of support is the highest that the government can
attain when the IFI does not assist but the interest group contributes to attain net
welfare V1: The IFI tenders an assistance schedule that makes the government
adopt policies such that the chosen assistance–distortion combination ðv1;ATÞ lies
on the GG curve. The fact that both ðv1;ATÞ and ðv2I ;A ¼ 0Þ lie on the GG curve
is described by Equation 10. The combination ðv1;ATÞ is determined by tangency
of the GG and I1I1 curves, where I1 expresses the IFI’s welfare at the conditional
assistance equilibrium. The tangency point reflects Equation 8. Clearly, the IFI is
better off with conditional assistance than no assistance. Without conditional
assistance, the IFI’s and, therefore, entire world’s welfare would be I0:Equations 8–10 can be solved to determine the government’s choice of
economic policies,v1; as well as the IFI’s and the interest group’s welfare, denotedby I1 and V1; respectively. Exogenous to the system are a and, indirectly, v2I :Since this chapter focuses on the efficiency of a one-time grant relative to loan
rollovers from the IFI’s point of view, this formulation has the special advantage
that it enables us to solve directly for the IFI’s welfare. Later, when we apply this
model to the case of “loans”, we will replace superscript “1” with superscript “L”.
12.3. INSTRUMENTS OF ASSISTANCE: LOAN ROLLOVERS VS.
A FINAL GRANT
When the IFI assists a country, it can do so by offering a loan or by making a grant.
This choice is available for unconditional, as well as conditional assistance. A loan
W. Mayer and A. Mourmouras260
provides assistance for a limited period of time. Its principal has to be repaid at a
specified time in the future. In addition, the assisted country might have to pay
interest. At the time of repayment, the IFI evaluates whether conditions that led to
the initial loan continue to exist or whether they have changed. A new loan might
be offered, whereby the new loan might be larger than, equal to, or smaller than
the original loan that is being repaid. Clearly, a major advantage of assisting a
country through loans is that it gives the IFI a great deal of flexibility. The IFI is
able to respond to changing benefits and opportunity costs of assisting, as well as
to a changing political climate in the receiving country.
A grant represents assistance that does not have to be repaid. Once the recipient
country is in possession of the grant, the IFI no longer can make adjustments. Even
if the developing country’s government changes its concern for the general
public’s welfare or the IFI’s opportunity cost of giving assistance rises, the grant
cannot be recalled. Clearly, this inability to adjust represents a drawback of
assisting through a grant rather than a loan. There are, however, advantages to
assisting through a grant. With no time schedule for repayment, a grant can be
utilized to finance both short- and long-term investment projects. Consequently,
the IFI’s inability to react to changing economic and political circumstances must
be weighed against the government’s enhanced ability to choose from a larger set
of investment projects.
The government’s ability to choose among projects and the IFI’s ability to
adjust the level of assistance are important influences on the IFI’s choice between
loans and grants. The purpose of this chapter is to highlight an important
additional aspect in the loans vs. grants debate, namely, the commitment aspect.
The provision of both unconditional and conditional assistance is the outcome of
interactions among three players: the policy-choosing government, the influence-
seeking domestic interest group, and the assistance-providing IFI. The nature of
the game these players engage in depends on whether assistance is unconditional
or conditional and whether it is given in form of a loan that is being rolled over or a
final grant.
This chapter highlights the differences in games within the framework of a
multi-period version of the model described in Section 12.2. The government of
the assistance-receiving country can adjust its policies only at the beginning of
each period. The adopted policy remains in place for the duration of this period. At
the beginning of the next period, however, the government can reassess and adjust
its policy. The domestic interest group also tenders its financial contribution
schedule at the beginning of each period, no matter whether the government
receives a loan or a grant. More precisely, at the beginning of each period, the
interest group presents its contribution schedule in the first stage of the game and
the government adopts a policy in the second stage of the game. The timing of the
Unconditional and Conditional Foreign Assistance 261
IFI’s assistance decision, on the other hand, depends on the form of assistance. It
is assumed that a loan is made available for the duration of one period only and
that no interest payments are charged. The value of each loan is determined at the
beginning of each period, and its equal-value repayment is required at the end of
the same period. In the case of an unconditional loan, the IFI’s decision is made at
the same (second) stage of the game as the government’s policy decision. In the
case of a conditional loan, the IFI presents its assistance schedule to the
government at the same (first) stage of the game as the interest group tenders its
financial contribution schedule.
A final IFI grant, on the other hand, represents a permanent commitment on the
part of the IFI not to interfere in the domestic political game after the conditions of
the grant have been met and the assistance has been disbursed. The final grant is
awarded at the beginning of the initial period only and it cannot be reversed
thereafter. If the grant is unconditional, the IFI decides at the same stage of the
game as the government; namely, in stage two of the initial period. At that time, it
does take account of the government’s reaction to this choice beyond the initial
period. If the final grant is conditional, the IFI’s grant schedule is tendered in stage
one of the initial period game, but the IFI again takes account of the government’s
responses in future periods. In particular, the IFI knows that a conditional grant
has no teeth beyond the initial period; if the government deviates from the
assistance schedule, the IFI has no power to recall the grant.
In order to highlight the commitment dimension of the final grant, we are going
to assume away all other distinctions between grants and loans. In other words, we
are going to specify a stationary, perfect foresight model in which economic and
political conditions are not expected to change from period to period.
Accordingly, there is no need for flexibility in awarding assistance over time.
Every period is like the initial period. In addition, the model assumes away all
distinctions between long- and short-term investments; they yield the same return.
Hence, in terms of investment returns, long-term funding through a final grant
offers no advantage over short-term funding through a loan rollover.
12.4. UNCONDITIONAL ASSISTANCE: THE IFI SHOULD USE
A GRANT
This section demonstrates that a final grant is superior to loan rollovers in pursuing
the IFI’s goal of maximizing world welfare when assistance is unconditional. This
conclusion is based on an evaluation of the impact of a loan that is being rolled
over indefinitely relative to a one-time grant when all players have perfect
foresight and no changes in political and economic conditions are foreseen. At the
W. Mayer and A. Mourmouras262
beginning of each period, the government makes its policy decisions for the
duration of the period. It does so under the influence of the domestic interest
group. The IFI also makes its loan assistance decision at the beginning of each
period; it makes its grant decision at the beginning of the initial period only.
12.4.1. Unconditional loan decisions
With no changes in the economic and political environment, the IFI faces the
same situation in each period when it decides on loan awards. It is dealing
with an incumbent government whose decisions are influenced by a domestic
interest group. The interactions between domestic government and interest
group have been laid out in Section 12.2.2. They were described by a non-
cooperative two-stage game in which the interest group tenders a political
contribution schedule, CðvÞ; in the first stage and the government makes its
policy choice in the second stage. The government chooses v such that
Equation 6 is satisfied. The IFI, in turn, offers a loan that maximizes world
welfare as defined in Equation 5. Since the IFI does not consider negative
loans to the developing country, A $ 0; the government’s policy choice, vL;and the IFI’s loan level, AL; must satisfy
U0ðvL
Þ2 ayðALÞ # 0 ð11Þ
ð12 vLÞy
0ðA
LÞ2 y
p 0ð2A
LÞ # 0 ð12Þ
where superscript “L” indicates equilibrium choices under a loan regime and
where Equations 11 and 12 hold as equalities for vL. 0 and AL
. 0;respectively. Since limv!0 U
0ðvÞ ¼ 1 and limv!1 U0ðvÞ ¼ 0; while yðALÞ . 0;
it must be that vL. 0: The value of AL; on the other hand, is positive or
zero. It is zero if, in the absence of assistance, the developing country’s
policies are so distorting that a transfer of resources to the developing country
lowers world output.
Figure 12.2 portrays a situation in which Equations 11 and 12 hold as equalities
and the equilibrium is unique. The RGRG curve is, as pointed out before, the best-
response curve of the domestic government based on Equation 11. It portrays the
government’s optimal choices of economic policies for all possible levels of IFI
assistance, given the interest group’s influence on political support for the
government. If there is no assistance, the government adopts policies that entail a
distortion index v2I : The RIRI curve is the best-response curve of the IFI based on
Equation 12. The IFI’s willingness to offer assistance declines with the magnitude
Unconditional and Conditional Foreign Assistance 263
of distortions in the assistance-receiving country. In equilibrium, distortion index
vL and assistance level AL prevail. At this “loan” equilibrium, marked by EL;welfare of the IFI is indicated by its assistance-distortion indifference curve ILIL:
12.4.2. Unconditional grant decisions
The IFI decides on awarding a final grant at the beginning of the initial period,
t ¼ 0: Since the grant is irreversible, it determines the stock of assistance capital
available to the domestic economy not only for the initial period, but for all
periods to come.7 The domestic government, on the other hand, makes its policy
decisions not just at the beginning of the initial period, but revisits it at the
beginning of each future period.
Starting with the recipient government, maximization of political support under
the influence of the interest group results in a distortion index vgt for each period
t ¼ 0; 1;…;1; such that
U0ðvg
t Þ2 ayðAtÞ # 0 ð13Þ
A
RG
IL
Ig
S
Ag
RI Eg
AL
EL
Ig
IL
RG
RI
O �g
�L
�-I
1 �
Figure 12.2: Superiority of unconditional grants.
7 It is implicity assumed that the capital created in the recipient economy with the assistance funds does
not depreciate over time.
W. Mayer and A. Mourmouras264
where superscript “g” indicates the best policy response when assistance in period
t; At; is received in form of a grant. Since, for a grant, A0 ¼ A1 ¼ · · · ¼ A; thegovernment chooses policies with the same distortion index for each period.
The IFI makes a grant decision only once, namely, at the beginning of the initial
period. It chooses a grant value, Ag; which maximizes the present value of world
output
½ð12 v0ÞyðAÞ þ ypð2AÞ� þ
X1
t¼1
dt½ð12 vtÞyðAÞ þ ypð2AÞ� ð14Þ
where 0 # d , 1 is the IFI’s discount factor. In the initial period, the value of
v0 is chosen simultaneously with A; in each future period t ¼ 1; 2;…;1; thevalue of vt is chosen once A is already in place. Accordingly, the IFI accounts
for the government’s future best responses to A; such that vt ¼ vðAÞ based on
Equation 13. The IFI’s present-value-maximizing choice of A; denoted by Ag;requires that
½ð12 v0Þy0ðA
gÞ2 y
p 0ð2A
g�
þ ½d=ð12 dÞ�{½12 vðAgÞ�y
0ðA
gÞ2 y
p0ð2A
gÞ2 yðA
gÞv0
ðAgÞ�} # 0
where v0ðAgÞ , 0: Recalling that v ¼ v0; the above equation can be reduced to
ð12 v0Þy0ðA
gÞ2 y
p0ð2A
gÞ2 dyðAg
Þv0ðA
gÞ # 0 ð15Þ
with equality holding for Ag. 0: The equilibrium values of the distortion
index, vg; and grant level, Ag; are attained from Equations 13 and 15 after
substitution of Ag for At in Equation 13 and of vg for v0 in Equation 15.
We now return to Figure 12.2 to compare the unconditional assistance
equilibrium under a grant, as described by Equations 13 and 15, with the
corresponding equilibrium under an infinite series of identical loans, as described
by Equations 11 and 12. The “loan” equilibrium occurs at point EL where RGRG
and RIRI ; the government’s and the IFI’s respective response functions, intersect.
The “grant” equilibrium must also lie on the government’s best-response curve,
RGRG; in order to satisfy Equation 13 as an equality. In addition, Equation 15
implies that, in equilibrium, the government’s best-response curve is flatter than
the IFI’s indifference curve. The slope of the government’s best-response curve is
1=v0; the slope of the IFI’s indifference curve is y=½ð12 vÞy0 2 yp 0�: Since d , 1;Equation 15 implies that the “grant” equilibrium, Eg; lies at a point between EL
and S: At point S; the IFI’s indifference curve is tangent to the government’s
Unconditional and Conditional Foreign Assistance 265
reaction curve. It would be the grant equilibrium if the IFI moved first in every
period, the initial one and all periods thereafter. In our model, however, the IFI
moves simultaneously with the government in the initial period and moves before
the government for all remaining periods. Accordingly, the more the future
counts, the larger the value of d; and the closer is Eg to point S:Figure 12.2 depicts two indifference curves of the IFI, ILIL and IgIg: The former
indicates the level of world welfare when the IFI assists through loans; the latter
expresses world welfare when the IFI provides assistance through a grant. Clearly,
the IFI, as the gatekeeper of world welfare, is better off along IgIg than along ILIL:Accordingly, it prefers to assist through a grant rather than through loans when
assistance is unconditional.
12.5. CONDITIONAL ASSISTANCE: THE IFI SHOULD USE LOANS
Economic assistance is conditional when it is contingent on the receiving
government’s adoption of specific economic policies. The IFI imposes the
conditions with the objective of raising world welfare. To be effective, conditional
assistance must avoid recidivism—i.e., it must lead to sustainable improvement in
economic policies. In the case of loans, achieving sustainable improvement in
policies is straightforward: if the assistance-receiving country deviates from the
conditions, the IFI does not renew the loan in the succeeding periods no matter
what the economic and political circumstances might be. Sustainable improve-
ment of economic policies is far more problematic if the IFI adopts a policy of
awarding a final conditional grant. A grant is an outright gift that permanently
moves all property rights of the transferred resources to the receiving government.
If the grant cannot be undone, then conditions imposed in the initial period cannot
be enforced in future periods. The IFI, therefore, anticipates the government to
adhere to the policy conditions for the initial period only and to switch over to
best-policy responses as soon as the government can revise its policy choice,
namely, as soon as the initial period is over.
12.5.1. Conditional loan decisions
The IFI’s loan decisions are made at the beginning of each period. With no
changes in economic and political conditions expected, the conditional loan
decision as well as the government’s choice of the distortion index will be the
same in each period. Our analysis, therefore, focuses on the conditions for a
truthful equilibrium during a given period only.
W. Mayer and A. Mourmouras266
The one-period conditional loan model is the same as the one-period model
examined in Section 12.2.3. It was set up as a two-stage game, in which the
interest group presents a contribution schedule and the IFI presents a loan
schedule in the first stage, while the government makes its policy choice in the
second stage. The conditions for a truthful equilibrium were stated as Equations
8–10. There, as here, it was implicitly assumed that no unconditional aid is
forthcoming. Equations 8–10 can be solved for the government’s conditional
policy choice in return for a loan, vL; as well as for the corresponding net utility ofthe interest group, VL; and IFI (world) welfare, IL in a given period. It is the IFI
welfare measure under a conditional loan, IL; that is of particular interest to us. We
want to compare it with the IFI welfare measure under a conditional grant, Ig;which will be discussed in Section 12.5.2.
12.5.2. Conditional grant decisions
The IFI awards a grant contingent on the government’s adoption of economic
policies that lower the distortion index to a specified value. The award is based on
a grant schedule that the IFI presents to the government at the beginning of the
initial period. The schedule spells out what size grant will be provided at all
possible initial distortion indices. By the nature of a grant, the chosen assistance
level remains the same for all periods to come. The government, on the other hand,
commits to a specific policy for one period only. Consequently, if it accepts
certain conditions for its policy choice in return for a given-size grant, it is bound
by these conditions only during the initial period. After the expiry of the initial
period, the government is free to choose those policies which, given the grant
received, maximize its political support. The IFI, of course, is aware of the
government’s recidivist incentives. Pursuing its best interest in the future leads the
government to backslide on the policies adopted when it accepted disbursement of
the one-time conditional grant.
For each period after the initial one, namely, periods t ¼ 1; 2;…;1;the government faces an inherited level of assistance capital, A; and chooses a
best-response distortion index vgt ; such that
U0ðvg
t Þ ¼ ayðAÞ ð16Þ
as was already stated in Equation 13. It follows that the same distortion
index prevails in all periods after the initial one and that vg ¼ vgt ¼ vðAÞ with
v0ðAÞ , 0:
Unconditional and Conditional Foreign Assistance 267
For the initial period t ¼ 0; on the other hand, the grant is conditional on the
adoption of IFI-prescribed policies. A truthful equilibrium requires that the
government chooses vg0; such that the present value of political support is
maximized;8 i.e.,
vg0 ¼ argmax
0#v,1
{CTðv0;V
g0 Þ þ að12 v0Þy½A
Tðv0; I
g�}
þ {d=ð12 dÞ}{CT½v;Vg
� þ a½12 v�y½ATðv0; I
gÞ�} ð17Þ
where v ¼ v½ATðv0; IgÞ�: In Equation 17, V
g0 and Vg denote equilibrium net
welfare of the interest group during the initial and all succeeding periods,
respectively. In addition, we use the symbol Ig ¼ {½12 d�½ð12 vg0Þy½A
Tðvg0; I
gÞ� þ
yp½2ATðvg0; I
gÞ�}þ {d½ð12 vgÞy½ATðvg0; I
gÞ� þ yp½2ATðvg0; I
g�} to indicate the
IFI’s per-period present value of welfare in equilibrium. Furthermore, the term
CTðv0;Vg0 Þ in Equation 17 states the initial-period truthful contribution schedule of
the interest group, and ATðv0; IgÞ is the corresponding truthful grant schedule of the
IFI. Finally, the term CT½vðATðv0; IgÞ;VgÞ� expresses the interest group’s truthful
contribution schedule for all periods beyond the initial one; during these periods, the
government’s policy choice is a response to the grant received in the initial period.
As shown in Appendix A, the choice of vg0 must satisfy
2ð12 dÞ{U0ðvg
0Þ2 ay½ATðvg0; I
g�}
aBy0½ATðvg0; I
gÞ�¼
›ATðvg0; I
gÞ
›v0
ð18Þ
where B ¼ ½ð12 vg0Þ2 dðvg
2 vg0Þ�: The left-hand side of Equation 18 states the
government’s willingness to accept assistance in return for lowering initial-period
distortions, with full realization that it will adopt best-policy responses beyond the
initial period. The right-hand term, on the other hand, expresses the IFI’s willingness
to award a grant in return for lowering initial-period distortions. The IFI is also aware
that the imposed policy conditions, willingly agreed upon by the government, are
binding only for the initial period, while the grant it has given stayswith the receiving
country forever. As shown in Appendix B, the right-hand side can be expressed as
›ATðvg0; IgÞ
›v0
¼ð12 dÞyðAgÞ
By0ðAgÞ2 yp0ð2AgÞ2 dv0ðAgÞyðAgÞð19Þ
where Ag ¼ ATðvg0; I
gÞ:
8 Note that the initial period’s choice of v0 indirectly influences later periods choices of vg:
W. Mayer and A. Mourmouras268
The interest group tenders its contribution schedule to the government in the first
stage of each period game. The contributionmust be sufficiently high to provide the
government with the same level of political support as it would receive if it did not
contribute. The equilibrium contributions differ between the initial period and the
ensuing periods. During the initial period, both interest group and IFI tender
payments schedules to the government. During the follow-up periods, only the
interest group presents a contribution schedule; the IFI’s payment is fixed, as
determined in the initial period. Maintaining the same level of political support for
the government during the initial period, t ¼ 0; and all remaining periods, t ¼
1; 2;…;1; respectively, requires that {CTðvg0;V
g0 Þ þ að12 vg
0Þy½ATðvg
0; IgÞ�} ¼
ay½ATð0; IgÞ� and {CTðvg;VgÞ þ að12 vgÞy½ATðvg0; I
gÞ�} ¼ ay½ATð0; IgÞ� which,in turn, implies that
Uðvg0Þ2 V
g0 þ að12 vg
0Þy½ATðvg
0; IgÞ� ¼ ay½AT
ð0; IgÞ� ð20Þ
UðvgÞ2 V
gþ að12 vg
Þy½ATðvg
0; IgÞ� ¼ ay½AT
ð0; IgÞ� ð21Þ
The IFI, on the other hand, tenders its grant schedule only once, namely, in the
first stage of the two-stage game that unfolds at the beginning of the initial period. It
also offers just enough to create the same present value of political support for the
government in equilibrium as the government would receive if no grant were
offered; in addition, the interest group’s utility must be retained at the same level as
in equilibrium. More precisely, the size of the grant must be such that
{CTðvg
0;Vg0 Þ þ að12 vg
0Þy½ATðvg
0; IgÞ�}þ ½d=ð12 dÞ�{CT
½vðATðvg
0; IgÞ;Vg
Þ�
þ a½12 vðATðvg
0; IgÞ�y½A
Tðvg
0; IgÞ}
¼ {CTðv2I ;Vg
0 Þ þ að12 v2IÞyð0Þ}þ ½d=ð12 dÞ�{CT
ðv2I ;VgÞ
þ að12 v2IÞyð0Þ�}
After substitution of CTðvg0;V
g0 Þ ¼ Uðvg
0Þ2 Vg0 and CTðvg;VgÞ ¼ UðvgÞ2 Vg;
this condition can be restated as
ð12 dÞ{Uðvg0Þ þ að12 vg
0Þy½ATðvg
0; IgÞ�}þ d{U½vðAT
ðvg0; I
gÞÞ�
þ a½12 vðATðvg
0; IgÞÞ�y½A
Tðvg
0; IgÞ�} ¼ Uðv2I
Þ þ að12 v2IÞyð0Þ ð22Þ
Equations 16, 18, and 20–22 constitute the conditions for a truthful equilibrium
when the IFI awards a conditional grant during the initial period and the government
knows that the IFI has no enforcement ability beyond this initial period.
Unconditional and Conditional Foreign Assistance 269
The system’s endogenous variables are the government’s policy choices, yielding
distortion indices vg0 and vg; respectively, for the initial and follow-up periods,
the corresponding net welfare levels of the interest group, Vg0 and V
g; and the IFI’sper-period welfare measure, Ig: It is the last of these variables in which we
have a particular interest. We want to compare it with IL; the per-period
measure of IFI welfare when assistance is given in form of loans for all periods to
come.
To make these IFI welfare comparisons, we first determine the Pareto-optimal
combination of distortion index,v; and grant level,A; given the constraint that jointwelfare of domestic government and interest group, Gþ V; is equal to what it is
in the absence of the IFI providing aid. Hence, we are choosingv andA tomaximize
I ¼ ½ð12 vÞyðAÞ þ ypð2AÞ� given the constraint ½UðvÞ þ að12 vÞyðAÞ� ¼½Uðv2IÞ þ að12 v2IÞyð0Þ�: Such optimal choice requires that
U 0ðvÞ2 ayðAÞ
að12 vÞy0ðAÞ¼
yðAÞ
yp0ð2AÞ2 ð12 vÞy0ðAÞð23Þ
Next, we look at the equilibrium conditions for rollover loans and compare
them with the Pareto-optimality condition. The conditions for the optimal choice
of vL and AL were stated in Equations 8 and 10, whereby we set vL ¼ v1 and
AL ¼ ATðv1; I1Þ: One can see immediately that the optimal distortion index
choice condition of Equation 8 is the same as the Pareto-optimality condition of
Equation 23. Also, the constraint on joint welfare of domestic government and
interest group under which the Pareto-optimal values of v and A were derived are
the same as the equilibrium values of distortion index and IFI per period assistance,
vL and AL: Accordingly, given the constraint of Equation 10, the loan equilibriumresults in the highest possible welfare per period for the IFI.
In case of a one-time grant, on the other hand, the condition for Pareto-
optimality cannot be satisfied. The constraint on joint welfare of domestic
government and interest group in the conditional grant model was stated on the
right-hand side of Equation 22. It is the same as the constraint in the loan model
and for the Pareto-optimal choices of v and A: If vg0 were equal to vðATÞ in
Equation 22, then the solution of the IFI grant level, ATð·Þ; would be the same as
the Pareto-optimal level of A: But since the initial-period government choice of
the distortion index, vg0; is less than its later period choice of vðATÞ—when the
grant conditions are no longer binding—it must be that Ag– A: Furthermore, the
government’s choice of the initial period, vg0; was shown to be the solution to
Equation 18. After substitution of Equation 19, this condition for the optimal
W. Mayer and A. Mourmouras270
choice of vg0 reduces to
{U 0ðvg0Þ2 ayðAgÞ}
aBy0ðAgÞ¼
yðAgÞ
yp0ð2AgÞ þ dv0ðAgÞyðAgÞ2 By0ðAgÞð24Þ
where again B ¼ ½ð12 vg0Þ2 dðvg
2 vg0Þ� and Ag ¼ ATðvg
0; IgÞ: Clearly, this
condition for the initial-period distortion index differs from the Pareto-optimality
condition due to the influence of B ¼ ½ð12 vg0Þ2 dðvg
2 vg0Þ� , ð12 vg
0Þ and of
dv0ðAgÞyðAgÞ , 0; assuming that the future matters ðd . 0Þ: Furthermore, the
government’s choice of the distortion index in later periods must be larger than
during the initial period as it no longer is constrained by IFI conditions.
Consequently, the award of a conditional grant implies that the adopted
combinations of (vg0; A
gÞ during the initial period and of ðvg; AgÞ during all
periods thereafter cannot be the same as the Pareto-optimal combination ðv;AÞ foreach period.
In deriving a specific Pareto-optimal combination of distortion index and IFI
assistance, ðv;AÞ; we imposed the condition that domestic government and
interest group are just as well off as in the absence of IFI assistance. The same
constraint is binding in determining the optimal combination of distortion index
and IFI assistance for conditional loan rollovers and for a conditional grant. For
the conditional loan rollovers, it was shown that ðv;AÞ is the choice in each
period. For a conditional grant, on the other hand, ðv;AÞ cannot be the choice ineach period. It follows that the maximizing value of IFI welfare under loans, IL;must be larger than the maximizing value of IFI welfare under a grant, Ig: Hence,in contrast to the grant bias under unconditional assistance, there exists a loan bias
under conditional assistance.
12.6. CONCLUDING REMARKS
The appropriate terms of IFI financial assistance to low-income countries,
including their length, interest rate and other repayment terms, are an important
policy concern for IFIs and the donor countries that provide these institutions with
their capital. To our knowledge, these issues have not been the subjects of
dynamic political economy analysis to date. We study some of these issues using a
dynamic version of the common agency model developed by Mayer and
Mourmouras (2002). In this model, the transfer of resources from the ROW to a
developing country has the potential to raise welfare of the world as a whole. It
thereby addresses the questions of how the world’s resources are best allocated
among countries but does not address temporary relief needs in crisis situations.
Unconditional and Conditional Foreign Assistance 271
The analysis confirms the inefficiency of prolonged financial association of IFIs
with low-income countries when assistance is unconditional. We demonstrate that
if there are no changes in either the ROW or the receiving country, it is best to
place unconditionally transferred resources permanently in the developing
country. If such a permanent resource transfer is called for and there is no
difference in returns between short- and long-run projects, then awarding a grant is
more efficient than employing renewable loans. The solution to the unconditional
assistance game is never Pareto-optimal. The IFI gains from using an initial period
grant rather than loans that are renewed every period since the grant gives the IFI a
first-mover advantage in its game with the government for all periods past the
initial one.
If, on the other hand, conditions are imposed on a policy-setting government,
then loan rollovers are more efficient than a one-time grant. Conditional loans can
be enforced from period to period, whereas the IFI’s enforcement capability of
grant conditions terminates after the initial period. Under a loan regime, the IFI is
able to set and enforce conditions that bring about a Pareto-optimal combination
of domestic economic policies and IFI assistance. Under a grant regime, the lack
of later-period enforceability makes it impossible to achieve a Pareto-optimal
outcome.
The question of the appropriate length of engagement of IFIs in developing
countries is closely related to the rationale for continued multilateral lending.
Our results suggest that for conditional assistance, permanent involvement of
international institutions is called for as long as domestic interest groups
resisting reforms remain organized in assistance-receiving countries. On the
other hand, limiting the time frame of unconditional assistance—subjecting IFI
assistance to a sunset clause—confers to the IFI a “first-mover’s advantage”.
When assistance is unconditional, this advantage enables an IFI that is active
for only one period to achieve better results than would be possible through
permanent engagement. There are two problems with this arrangement,
however. One is that it fails to achieve a Pareto-optimal allocation for the
world economy. While the transfer of resources results in a permanent
improvement in policies relative to the no-assistance equilibrium, policies
worsen after the IFI withdraws and reform-resisting interest groups reassert
themselves. The second weakness is that final grants are not time consistent. To
implement the final grant policy, the IFI must commit to not reengage in the
assistance-receiving country. But unless the interest groups lobbying for
distortionary policies become a spent force under the final-grant policy, the IFI
will face incentives to re-engage. These incentives will be stronger when the
unexploited gains from the political game between the IFI and the assistance-
receiving country are larger.
W. Mayer and A. Mourmouras272
As is true in all theoretical investigations, our results are derived within a
model that abstracts from a variety of considerations that are likely to be
important in practice. In the real world, the opportunity costs of giving aid, the
benefits from receiving aid, and the political concern of the receiving
government for its general public undergo frequent changes. Consequently,
all decision makers face an uncertain future and the IFI will find it
advantageous to employ loans instead of grants to adjust to changing
circumstances. The IFI also will be aware that the length of time for which it
transfers resources affects the type of investment project to which these
resources are directed. Generally, a long-term resource transfer offers more
flexibility with respect to investment projects than a short-term one. Accounting
for both of these influences, the IFI attempts to find the right mixes of loans and
grants. Given this context, the message of this chapter is that, in the real world,
the IFI must also account for a third influence on its decision, namely, its ability
to influence the behavior of the assistance-receiving government. Under
unconditional assistance, the grant commits the IFI and enables it to benefit
from its first-mover situation. Under conditional assistance, conditionality has
the potential to attain Pareto-optimality. But this works only if conditions are
enforceable in every period, as they are for loans but not for a grant.
The vagaries faced by developing countries suggest the need for permanent
institutions to assist countries that fall into poverty at different random times
(either because of exogenous shocks or other reasons beyond their control). To
assist low-income countries, IFIs need to be endowed with assistance funds before
crises erupt. Future research should determine the modalities of IFI assistance to
low-income countries taking into account both the need for flexibility in the
provision of assistance and the moral hazards of prolonged financial association of
IFIs in developing countries. The answer to this question, which has an analogue
in the design of national safety nets to deal with household income and other risks,
will require more complicated models but is a key issue for thinking about the
replenishment of funds of assistance-providing IFIs.
The dynamic version of the lobbying model developed in this chapter suggests
two additional methodological topics for future research. The first is the question
of whether IFI-supported structural reforms contribute to a sustainable decline in
the pernicious influence of special interest groups in assistance-receiving
countries. This question, in turn, is key to addressing the factors blocking the
emergence of strong institutions and improving the business climate. In our
framework, the organized interest group is a permanent fixture of the institutional
landscape in the assistance-receiving country. While its position is weakened as a
result of IFI operations and its equilibrium rents and political influence decline,
the interest group does not go away. In reality, of course, interest groups are able
Unconditional and Conditional Foreign Assistance 273
to organize themselves only if they can overcome the free rider and other
collective action problems as argued by Olson (1965). It would be interesting to
study how the survival of interests groups is affected by IFI conditionality if there
are costs to setting them up as in Mitra (1999).
In a model in which interest groups become organized only if their members
pay up some fixed sunk cost, the possibility arises that IFI involvement could push
the rents earned by interest groups below the threshold required for organized
lobbying activity to be profitable. If the IFIs knew the threshold below which
the organized interest group is “disarmed,” they could then strategically
manipulate their conditional assistance to drive the interest group’s profits
below this critical value, resulting in permanent, long-term improvements in the
recipient country’s institutional climate. The provision of assistance aiming to
destroy the interest group as an organized force could still be modeled as a truthful
equilibrium, since the welfare benefits from interest group “disarmament” are
real. Such assistance could be compared to the possibility analyzed in the present
chapter, namely, that rents are higher than the fixed cost of organizing the interest
group, in which case the interest group continues to operate but is permanently
worse off as a result of the IFI’s intervention. This idea provides a natural link
between the lobbying literature and the literature on the new institutional
economics (see North, 1993) and could aid in formulating a positive theory of
institutional evolution.
The second methodological issue concerns the need for models of IFI-
recipient country interactions that incorporate explicit public choice dynamics.
IFIs increasingly provide their assistance to democratic governments that are
subject to electoral dynamics (and street protests and other forms of political
influence) when they reorient resources from non-traded to traded goods
industries and undertake other reforms. IFIs are often surprised by the political
dynamics associated with the reform programs they support, including those
around elections. A better understanding of such dynamics would help in the
design of politically sustainable reform programs (Willett, 2003). In our theory,
the equilibrium level of political support for recipient governments is
endogenously determined over time, yet incumbents never get thrown out of
office in equilibrium. This outcome reflects the lack of voting or other explicit
public choice mechanism in the version of the Grossman and Helpman theory
we employ. Extensions in this direction seem feasible. Grossman and Helpman
(1996) have extended their model to rationalize their choice of objective
function. Analyzing the impact of IFI assistance on an endogenously
determined domestic political process is a promising avenue for future
research.
W. Mayer and A. Mourmouras274
ACKNOWLEDGEMENTS
We would like to thank Burkhard Drees and Andy Feltenstein for useful
comments on a previous version. All remaining errors are our own.
APPENDIX A. DERIVATION OF EQUATION 18
Since CTðv0;Vg0 Þ ¼ Uðv0Þ2 V
g0 and CTðv;VgÞ ¼ U½vðATðv0; I
g�2 Vg; differ-entiation of Equation 17 with respect to v0 implies the first-order condition
{U 0ðvg
0Þ2 ayðAgÞ þ að12 vg
0Þy0ðA
gÞð›A
g=›v0Þ}þ {½d=ð12 dÞ�{½U 0ðvg
Þ
2 ayðAgÞv0
ðAgÞð›A
g=›v0Þ þ a½12 vðAgÞ�y
0ðA
gÞð›A
g=›v0Þ} ¼ 0
where Ag ¼ ATðvg0; I
gÞ: Since U 0ðvgÞ ¼ ayðAgÞ from Equation 16 due to the
government’s optimal response to the grant past the initial period, one can solve
the first-order condition for
2ð12 dÞ½U 0ðvg
0Þ2 ayðAgÞ�
ay0ðAgÞ½ð12 vg0Þ2 dðvg 2 vg
0Þ�¼
›Ag
›v0
APPENDIX B. DERIVATION OF EQUATION 19
The term ATðvg0; I
gÞ denotes the IFI’s truthful offer of a grant when the
government chooses the equilibrium initial-period distortion index and the IFI’s
equilibrium welfare is Ig: It is derived from the definition of Ig ¼ ð12 dÞ �½ð12 vg
0ÞyðAgÞ þ ypð2AgÞ� þ d{½12 vðAgÞ�yðAgÞ þ ypð2AgÞ}: At a given Ig;
›Ag
›v0
¼ð12 dÞyðAgÞ
y0ðAgÞ½ð12 vg0Þ2 dðvg 2 vg
0Þ�2 yp 0ð2AgÞ2 dv0ðAgÞyðAgÞ
REFERENCES
Bernheim, B. D. and Whinston, M. D. (1986). Menu auctions, resource allocation, and
economic influence. Quarterly Journal of Economics, 101, 1–32.
Bird, G., Hussain, M. and Joyce, J. (2000). Many happy returns? Recidivism and the IMF.
Wellesley College Department of Economics Working Paper, 2000-04.
Unconditional and Conditional Foreign Assistance 275
Cunningham, S. (2002). G7 finance ministers end row over world bank grants, loans.
Reuters North American Securities News (June 15).
Dixit, A., Grossman, G. M. and Helpman, E. (1997). Common agency and coordination:
general theory and application to government policy making. Journal of Political
Economy, 105, 752–769.
Drazen, A. (2002). Conditionality and ownership in IMF lending: a political economy
approach. IMF Staff Papers, 49, 36–67.
Friedman, M. (1958). Foreign economic aid: means and objectives. Yale Review, 500–516.
Grossman, G. M. and Helpman, E. (1994). Protection for sale. American Economic Review,
84, 833–850.
Grossman, G. M. and Helpman, E. (1996). Electoral competition and special interest
politics. Review of Economic Studies, 63, 265–286.
International Monetary Fund (2002). Evaluation of prolonged use of IMF resources.
Independent Evaluation Office, vols. 1 and 2. Available via the Internet at http://www.
imf.org/external/pubs/ft/EPUI/2002/pdf/Report.pdf.
International Monetary Fund (2003a). Conclusions of the task force on prolonged use of
fund resources, (February 4). Available via the Internet at http://www.imf.org/external/
np/pdr/ufr/2003/020403.pdf.
International Monetary Fund (2003b). Operational guidance for assessments of countries
with a longer-term program engagement (August 20). Available via the Internet at http://
www.imf.org/external/np/pdr/ufr/2003/082003.htm.
International Monetary Fund (2003c). Fund assistance for countries facing exogenous
shocks (August 8). Available via the Internet at http://www.imf.org/external/np/pdr/
sustain/2003/080803.pdf.
Mayer, W. and Mourmouras, A. (2002). Vested interests in a positive theory of IFI
conditionality, IMF Working Paper, Washington, DC: International Monetary Fund.
Mitra, D. (1999). Endogenous lobby formation and endogenous protection: a long run
model of trade policy determination. American Economic Review, 89, 1116–1134.
North, D. (1993). Economic performance through time. Nobel Memorial Prize Lecture,
Stockholm (December 9). Available via the Internet at http://www.nobel.se/economics/
laureates/1993/north-lecture.html.
Olson, M. (1965). The Logic of Collective Action, Cambridge: Harvard University Press.
Willett, T. (2003). IMF conditionality and the new political economy of ownership,
Working Paper, Claremont, CA: Claremont McKenna College.
W. Mayer and A. Mourmouras276
CHAPTER 13
Crowding Out and Distributional
Effects of FDI Policies
AMY JOCELYN GLASSa,* and KAMAL SAGGIb,1
aDepartment of Economics, Texas A&M University, Allen Building,
College Station, TX 77843-4228, USAbDepartment of Economics, Southern Methodist University, 3300 Dyer Street, Suite 301,
Dallas, TX 75275-0496, USA
Abstract
To examine the effects of international investment agreements, suppose firms
from two source countries invest in a host country. An increase in the tax faced by
firms from one source decreases foreign direct investment (FDI) from that source
and increases FDI from the other. When free to discriminate, the host country
imposes a larger tax on multinationals that benefit more from FDI. The source
country whose multinationals pay the larger tax would gain from a most-favored
nation standard requiring non-discrimination relative to other foreign firms, and
would gain further from a national treatment standard requiring non-discrimi-
nation relative to domestic firms.
Keywords: Foreign direct investment, taxes, subsidies, oligopoly
JEL classifications: F1, F2, L1
*Corresponding author. Tel.: þ 1-979-845-8507; fax: þ 1-979-847-8757.
E-mail addresses: [email protected] (A.J. Glass), [email protected] (K. Saggi)1 Tel.: þ 1-214-768-3274; fax: þ 1-214-768-1821.
13.1. INTRODUCTION
Foreign direct investment (FDI) occurs in an environment in which countries are
relatively free to pursue investment policies of their own choosing. While trade
policies of member countries of the World Trade Organization (WTO) are subject
to many disciplines, investment policies presently are not.2 Whether the member
countries of the WTO should negotiate a multilateral investment agreement is an
unresolved policy issue—see Hoekman and Saggi (2000) for an overview.
In the absence of a multilateral agreement on investment, bilateral
investment treaties have proliferated. Two common standards of treatment
used in these treaties are most-favored-nation (MFN) and national treatment.
MFN requires non-discrimination relative to other foreign firms, while national
treatment requires non-discrimination relative to domestic firms. The MFN
standard is far closer to being generally accepted than the national treatment
standard (UNCTAD, 1999a). Most host countries still adopt policies that favor
domestic firms.
Despite the practical importance of the issue, the literature largely ignores how
FDI policies alter production location decisions and impact firms and workers in a
multicountry setting.3 When the literature does move beyond a two-country
setting, it focuses on multiple host countries competing for FDI from a source
country.4
To emphasize the third country repercussions of FDI policies, we develop a
model where a host country sets its policy toward FDI from multiple source
countries. A model with one host and two source countries is the simplest scenario
for considering issues related to third country repercussions and non-discrimi-
nation requirements. We model FDI as occurring in oligopolistic markets,
consistent with the observations made by Brainard (1997) and Markusen (1995).
FDI shifts labor demand across countries and thus affects wages and profits.
Consequently, the host country’s policy toward FDI from one country has
repercussions not only for the extent of FDI from the alternative source country
2 The Uruguay round did lead to the agreement on Trade Related Investment Measures (TRIMS) that
came into effect in 1995. However, this agreement deals only with those investment policies that are
trade-related, such as domestic content requirements and licensing requirements. Investment policies
face no direct disciplines. Furthermore, even some key TRIMS such as export performance
requirements are not covered by this agreement.3 See Horstmann and Markusen (1989, 1992), Janeba (1996, 1998), Markusen and Venables (1998,
1999), Motta (1992), Rowthorn (1992), and Raff and Srinivasan (1998).4 Haufler and Wooton (1999) examine choice of a lump-sum profit tax by two competing host
governments. Motta and Norman (1996) argue that regional integration can attract FDI and lead to
FDI incentives.
A.J. Glass and K. Saggi278
but also for wages and profits in all countries. In particular, we show that FDI
policies alter the
† composition of FDI across source countries (crowding out effect),
† distribution of rents between firms (strategic effect),
† distribution of income between workers and firms (distributional effect).
We characterize the optimal FDI policy profile offered by the host government:
the tax or subsidy on multinationals from each source country. Inward FDI raises
wages thereby benefiting workers and hurting host firms. This tension between
the interests of the two groups determines optimal host policy.We find that the host
government levies a higher tax on multinationals from the source country with the
smaller labor supply per firm and thus the stronger natural tendency to conduct FDI.
What are the consequences of requiring the host country to abide by the MFN
principle—that the host country must tax or subsidize all multinational production
to the same degree? Clearly, the host country is harmed by any restriction on its
freedom to set policies differentially. But the disfavored source country benefits
from equal treatment through a reduction in the tax imposed on its multinationals
and thus has an incentive to push for MFN treatment. Likewise, the disfavored
source country also would benefit from requiring national treatment—that the host
country must tax or subsidize production by multinationals no differently than
production by host firms.
After setting up the model, we examine the properties of the no intervention
equilibrium. Then we consider FDI policies adopted by the host country and the
effects of imposing non-discrimination constraints on these policies. We also
comment on the case where firms from different countries have different
technologies of production, thereby generating another basis for policy
discrimination.
13.2. MODEL
The world consists of three countries indexed by i : a host country receiving FDI
from two source countries i ¼ 1; 2:5 Each country i has n symmetric oligopolistic
industries comprised of mi number of firms. There is one factor of production,
skilled labor; any reference to labor should be understood to mean skilled labor.
Producing one unit of output of any good requires one unit of skilled labor in any
country.
5 Any fixed cost of FDI has already been paid by each source firm.
Crowding Out and Distributional Effects of FDI Policies 279
Source firms decide whether to produce each unit of output at home or abroad.
Let ai denote the share of skilled labor demanded in the host country by a firm
from source country i; which provides a measure of the extent of FDI in the host
country from source country i: Figure 13.1 illustrates the FDI flows considered in
our model.
Host country policy alters the incentives for FDI. Suppose that a firm from
country i faces an output tax of ti for each unit of output produced in the host
country.6 From the perspective of a firm from a source country, the tax increases
the marginal cost of producing in the host country. If source firms split production
across countries, 0 , ai , 1; the marginal cost of production (including any tax
or subsidy) must be equalized across countries
zi ¼ z0 þ ti; ;i ¼ 1; 2; ð1Þ
where zi denotes the wage in source country i and z0 denotes the wage in the host
country. Each firm views the wage in each country as given as there are many
firms hiring skilled labor in each country.
Firms behave as Cournot oligopolists. The demand function facing the world
industry is given by P ¼ pðQÞ where p0ðQÞ , 0 and p00ðQÞ # 0: Let yi denote theoutput of a firm from country i. Total world output equalsQ ;
P2i¼0 miyi: Profit of
a firm in country i is pi ¼ ðp2 ciÞyi; where ci ¼ aiðz0 þ tiÞ þ ð12 aiÞzi for
source firms i ¼ 1; 2 and c0 ¼ z0 for host firms. Profit maximization requires the
first order conditions
p2 yic ¼ ci; ;i ¼ 0; 1; 2; ð2Þ
Figure 13.1: FDI pattern with two source countries.
6 Since taxes are not restricted to be positive, we allow subsidies. Other policy instruments, such as
profit taxes and local content requirements, are common, especially for restricting FDI—see Lahiri and
Ono (1998). Profit and output taxes impact production location decisions and wages in a similar
fashion.
A.J. Glass and K. Saggi280
where p ; pðQpÞ . 0 is the price and c ; 2p0ðQpÞ . 0 is the negative of the
slope of the demand function at the equilibrium industry output. Applying the FDI
equilibrium conditions to simplify the first order conditions yields p2 yic ¼ zi as
ci ¼ zi for any ai . 0:We examine an environment where output in each country is constrained by the
supply of skilled labor. Denote the skilled labor supply per industry in country i by
ki (with K ;
P2i¼0 ki as the total labor supply per industry).7 Since only a fixed
supply of workers are available in each country i, wages adjust to clear labor
markets. The labor market equilibrium condition for the host country is
m0y0 þX2
i¼1
aimiyi ¼ k0; ð3Þ
where labor demand in the host country equals host firm production plus
multinational production from both source countries.8 Similarly, the labor market
equilibrium conditions for the two source countries are
ð12 aiÞmiyi ¼ ki; ;i ¼ 1; 2; ð4Þ
where labor demand in each source country is the share of multinational
production kept in the source country. Adding together the labor constraints
(Equations 3 and 4) yields that total output is constrained by the total availability
of labor Qp ¼ K and is hence unresponsive to changes in FDI policies.
Define welfare in the host country as the sum of profit and labor earnings plus
any tax revenues (or minus any subsidy payments)
W0 ¼ Bm0p0 þ z0k0 þ T ; ð5Þ
where 0 # B # 1 denotes host country ownership share of host firms and total tax
revenues are
T ¼ t1a1m1y1 þ t2a2m2y2: ð6Þ
Define welfare in each source country as the sum of profit and labor earnings,
Wi ¼ bm0p0 þ mipi þ ziki; ;i ¼ 1; 2; ð7Þ
7 We maintain the notation of Dixit and Grossman (1986) and Glass and Saggi (1999), who refer to
skilled labor as k.8 Brander and Spencer (1987), Bughin and Vannini (1995), and Das (1981) consider models with
unemployment in the host country.
Crowding Out and Distributional Effects of FDI Policies 281
where 0 # b # 1 denotes source country ownership share of host firms and
B ; 12 2b generates full ownership of host firms within these three
countries.9
Define world welfare as the sum of host and source countries welfare W ;
W0 þW1 þW2: We measure welfare net of consumer surplus since consumer
surplus does not respond to policy changes due to total output being fixed.10 In our
model, policy intervention alters only the distribution of welfare across countries.
This feature is useful in highlighting strategic and distributional effects of FDI
policies, the focus of this chapter.
13.3. NO INTERVENTION AND NATIONAL TREATMENT
An equilibrium specifies the output of firms {y0; y1; y2}; the wage in each country
{z0; z1; z2}; and the extent of FDI from each source country {a1;a2}: Let
{yni ; zni ;a
ni } denote the optimal solution to the three first order conditions
(Equation 2), the three labor constraints (Equations 3 and 4) and the two FDI
equilibrium conditions (Equation 1) when t1 ¼ t2 ¼ 0:In the absence of government intervention, the equilibrium extent of FDI from
source country i into the host country is
ani ¼ 12
M
mi
ki
K; ;i ¼ 1; 2; ð8Þ
where M ;
P2i¼0 mi the total number of firms in the world. Label the source
countries so that the first source country has a larger labor supply per firm than the
second source country.
k0
m0
.k1
m1
$k2
m2
: ð9Þ
To ensure that FDI does indeed occur from the first source country an1 . 0; we
additionally assume that the first source country has fewer resources per firm than
9 We assume source firms are fully owned within the source country (and the source countries own the
same share of host firms) for simplicity. Allowing the source countries to own different shares of host
firms would not affect the policies chosen by the host country.10 An absence (or lack of importance) of consumer surplus effects may also occur for export-oriented
FDI, where sell (primarily) to an external market. Such a situation is particularly relevant when FDI
occurs to lower production costs, as is the case in our model.
A.J. Glass and K. Saggi282
the world average
k1
m1
,K
M: ð10Þ
This assumption for the first source country is stronger than the ordering due to the
labeling of countries (Equation 9).
In our model, FDI arises due to labor scarcity in the source country relative to
the host country (or world). The equilibrium extent of FDI from a source country
is smaller the larger the labor supply in the source country relative to the world.
The intuition is that a larger labor supply implies a smaller incentive for FDI.
Similarly, the equilibrium extent of FDI from a source country is larger the larger
the number of firms in the source country relative to the world. Proofs appear in
Appendix A.
Proposition 1. An increase in the labor supply of a source country relative to
the world decreases the extent of FDI from that source country. An increase in the
number of firms in a source country relative to the world increases the extent of
FDI from that source country.
Define di ; ki=k0 as the labor supply of source country i relative to the host
country. Figure 13.2 illustrates the FDI patterns in ðd1; d2Þ space. From Equation
8, the line D1 depicts the boundary for FDI to occur from the first source
country
d2 . 21þM
m1
2 1
� �
d1 , an1 . 0; ð11Þ
Figure 13.2: Equilibrium FDI pattern.
Crowding Out and Distributional Effects of FDI Policies 283
and the line D2 depicts the boundary for FDI to occur from the second source
country
d2 ,1
M
m2
2 1
ð1þ d1Þ , an2 . 0: ð12Þ
The area where firms from both source countries invest in the host country
ðan1 . 0 and an
2 . 0Þ lies between the two lines (below D2 and above D1). In
this region, the labor supply in each source country is small relative to the host
country.
In the absence of government intervention, FDI equalizes wages across
countries11
zni ¼ z
n¼ p2
cK
M; ;i ¼ 0; 1; 2: ð13Þ
In equilibrium, the wage in any country decreases with an increase in labor in any
country and increases with an increase in the number of firms in any country. A
greater number of firms generates greater labor demand, which elevates the wage;
a greater labor supply depresses the wage.
The output of each firm reflects the average labor supply per firm in the world
yni ¼ y
n¼
K
M; ;i ¼ 0; 1; 2: ð14Þ
The output of a firm from any country clearly increases with an increase in the
labor supply in any country, and decreases with an increase in the number of firms
in any country (as the constant total output must be split across a larger number of
firms).
The driving force behind these results is that any change (a decrease in the labor
supply or an increase in the number of firms relative to the world) that increases
the wage in a source country relative to the host country encourages more FDI in
order to restore the equality of wages across countries. When the labor supply in
the first source country increases relative to the world, its firms shift less
production abroad.
The structure of our model makes only the tax difference relative to host firms
matter for the location of production and a country’s welfare. We set the tax on
host firms to zero. Due to the fixed labor supply, any common tax on production
11 Wages are measured in efficiency units of labor, so observed wages may still differ across countries.
A.J. Glass and K. Saggi284
by all firms manifests itself only in the host wage falling by the amount of the tax
so that the unit production cost inclusive of the tax is unchanged—see Dixit and
Grossman (1986) for a similar feature. Output, FDI, and welfare levels are
unaffected. Hence we can view our nonintervention equilibrium as representing
national treatment, where multinationals are taxed to the same degree as are local
firms. The tax on multinational production in what follows should, therefore, be
viewed as the tax on multinationals in excess of the tax on local firms (and wages
as net of the tax on local firms).
13.4. DISCRIMINATORY TREATMENT
Consider the scenario where the host government can set FDI policies that differ
across firms from different countries (that is, the host country violates the MFN
principle). We describe equilibrium under such discriminatory intervention and
then investigate optimal policies.
13.4.1. Equilibrium
Let {ypi ; zp
i ;ap
i } denote the optimal solution to the three first order conditions
(Equation 2), the three labor constraints (Equations 3 and 4) and the two FDI
equilibrium conditions (Equation 1). The equilibrium extent of FDI from each
source country is
ap
i ¼
mjtj þ c K 2M
mi
ki
� �
2 tiðM 2 miÞ
mjtj þ cK 2 tiðM 2 miÞ: ð15Þ
The extent of FDI from the first source country decreases with its own FDI tax and
increases with its rival’s FDI tax; the extent of FDI from the second source country
has identical properties.
Equilibrium wages in the host country equal
zp
0 ¼ p2cK þ m1t1 þ m2t2
M; ð16Þ
and equilibrium wages in each source country equal
zp
i ¼ p2cK 2 ðM 2 miÞti þ mjtj
M; ;i; j ¼ 1; 2; i – j: ð17Þ
Crowding Out and Distributional Effects of FDI Policies 285
Wages decrease in each country with an increase in the world labor supply. The
host country wage decreases with either FDI tax; the wage in each source country
rises with its own FDI tax and falls with its rival’s FDI tax. This latter effect is
interesting as it indicates that a tax on FDI from source country i impacts wages in
rival source country j.
Lastly, equilibrium output of a local firm in the host country is
yp
0 ¼K
Mþ
m1t1 þ m2t2Mc
; ð18Þ
while equilibrium output of a multinational firm from each source country is
yp
i ¼K
Mþ
mjtj 2 tiðM 2 miÞ
Mc; ;i; j ¼ 1; 2; i – j: ð19Þ
Equilibrium output of a multinational firm falls with its own FDI tax and rises with
its rival’s FDI tax. Equilibrium host firm output increases with the tax on FDI from
either source country. In our next proposition, we consider the effects of raising
the tax on FDI from one of the source countries; a full analysis of the policy
equilibrium follows in Section 13.4.2.
Proposition 2. A tax on multinational production by firms from one source
country discourages FDI from the source country whose firms are subject to the
tax while encouraging FDI from the other source country, increases wages in the
disfavored source country while decreasing wages both in the other source
country and in the host country, and expands output (and profits) for firms from
the favored source country and host firms, while contracting output (and profits)
for firms from the other source country.
The above proposition demonstrates the third country effects of FDI policies.
Since national labor markets are linked by FDI in our model, the host country’s
FDI policy has repercussions not only for the FDI flows from the alternative
source country but also for the wages (and therefore, the output levels and profits
of firms) in both countries.
13.4.2. Policy
We calculate welfare in each country by substituting the equilibrium values of the
endogenous variables from Section 13.4.1. An increase in the FDI tax on the
A.J. Glass and K. Saggi286
output of firms from the first source country affects the host country’s welfare
according to dW0=dti: Setting dW0=dt1 ¼ 0 and likewise dW0=dt2 ¼ 0
determines the optimal FDI policy schedule.
ti ¼c
2
K þ k0M
B2 mi 2 mj
2ki
mi
2
664
3
775; ;i; j ¼ 1; 2; i – j: ð20Þ
Firms from both source countries are subject to FDI taxes ðtp1 . 0 and tp2 . 0Þ
provided world labor supply (host labor supply) and host country ownership of
host firms are not too small.12
Figure 13.3 illustrates the range of relative labor supplies di ; ki=k0 where
firms from both source countries suffer FDI taxes from the host government. For
each source country, its relative labor supply must be sufficiently small to suffer a
tax. The boundary lines are the line D3 for firms from the first host country to
suffer a tax on their FDI
d2 . 22þ2m1 þ m2 2
M
Bm1
0
BB@
1
CCAd1 , tp1 . 0; ð21Þ
Figure 13.3: Equilibrium FDI policy.
12 If the host country government cares strongly about local wages (B is small), it may subsidize inward
FDI. This result suggests that the use of incentives to attract FDI may stem from the desire to improve
the welfare of workers in host countries.
Crowding Out and Distributional Effects of FDI Policies 287
and the line D4 for firms from the second host country to suffer a tax on their FDI
d2 ,m2
M
B2 ðm1 þ 2m2Þ
0
BB@
1
CCAð2þ d1Þ , tp2 . 0: ð22Þ
Thus, firms from both countries suffer taxes ðtp1 . 0 and tp2 . 0Þ when the relative
labor supplies of both countries are sufficiently small. There also exist ranges
where FDI from only one of the source countries is taxed (while FDI from the
other is subsidized) and where FDI from both source countries is subsidized.
Comparing the taxes on firms from the two countries, the optimal tax schedule
implies the difference
Dt ¼ tp1 2 tp2 ¼c
2
k2
m2
2k1
m1
� �
: ð23Þ
The optimal tax is lower for the first source country than the second tp1 , tp2because the first source country has a larger labor supply per firm than the second
(Equation 9), making the firms from the first country more sensitive to taxes
on FDI.
Proposition 3. Firms from the source country with the larger labor supply
per firm and hence the smaller natural tendency to conduct FDI face a smaller tax
on multinational production.
13.4.3. Discriminatory versus national treatment
Let DWpni ; Wp
i 2Wni be the extent that welfare in country i is higher under
discriminatory FDI policies than national treatment. Define Dk ; k1 2 k2: For thesame number of firms mi ¼ m; the first source country gains more from
discriminatory FDI policies relative to national treatment than the second source
country due to its larger labor supply.
DWpn12 ; DWpn
1 2 DWpn2 ¼
cKDk
4M1þ 3
k0
K
� �
. 0: ð24Þ
The host country must gain: it can always set the FDI taxes to zero so it can do no
worse being able to set its FDI policy. World welfare is fixed, so one source
A.J. Glass and K. Saggi288
country must lose. We find that the second source country necessarily loses from
discriminatory (relative to national) treatment while the first source country might
gain or lose. Hence, the second source country, the one that suffers the larger tax,
would have an incentive to push for national treatment. This result also holds true
if multinational production is subsidized: the second source country gets the
smaller subsidy and prefers that all subsidies be eliminated. Next, we check
whether the second source country benefits from MFN treatment, and whether it
still benefits from national treatment if it already enjoys MFN treatment.
13.5. MOST-FAVORED-NATION TREATMENT
Now we consider host country intervention, but where the tax on FDI is
constrained to be the same across firms from different source countries t1 ¼ t2 ¼t: This scenario represents MFN treatment, where a foreign firm cannot be treated
worse than the best that any other foreign firm is treated.
13.5.1. Equilibrium
Let {ysi ; zsi ;a
si} denote the optimal solution to the three first order conditions
(Equation 2), the three labor constraints (Equations 3 and 4) and the two FDI
equilibrium conditions (Equation 1) when t1 ¼ t2 ¼ t: Equilibrium expressions
for the endogenous variables appear in the Appendix A.
Proposition 4. A symmetric tax on multinational production reduces FDI as
well as the output (and profits) of firms from both source countries, increases profit
and output of host firms, and raises wages in both source countries while lowering
wages in the host country.
The above proposition highlights the strategic as well as distributional
consequences of FDI policies. A tax on FDI reduces the wage in the host country
and thus decreases the marginal cost of host firms thereby increasing their profits
at the expense of source firms.
13.5.2. Policy
We calculate welfare in each country by substituting the equilibrium values of
the endogenous variables. An increase in the symmetric FDI tax affects the host
country’s welfare according to dW0=dt: Setting dW0=dt ¼ 0 determines
Crowding Out and Distributional Effects of FDI Policies 289
the optimal FDI tax under the MFN principle.
t s¼
c
2
K þ k0M
B2 ðM 2 m0Þ
2K 2 k0
M 2 m0
2
664
3
775: ð25Þ
Increases in the labor supply (or decreases in the number of firms) in either source
country lower the tax level. The FDI tax is positive provided the world labor
supply (implicitly the host labor supply) and weight on host firm profits are not too
small
dW0
dt. 0 ,
k1 þ k2
m1 þ m2
, 2BK
M: ð26Þ
This expression is the counterpart to D3 and D4 but for symmetric FDI policy.
13.5.3. Discriminatory versus MFN treatment
A comparison of the MFN tax with the discriminatory taxes indicates how the
policies adjust under mandated symmetric treatment.
Proposition 5. If forced to treat firms from the two source countries
symmetrically, the host country increases the FDI tax on firms from the larger
country and decreases theFDI tax on firms from the smaller country: tp1 , t s, tp2:
The consequences of this policy change are clear: FDI from the favored source
country is squeezed out if discriminatory treatment is replaced by MFN treatment.
Such a policy change also has distributional consequences because firms from the
first country obviously lose (and its workers gain) when their tax increases
whereas firms from the second country gain (and its workers lose). Thus firms
from the source country with the lower tax would object to any move toward equal
treatment of multinationals.
Let DW spi ; W s
i 2Wp
i be the extent that welfare in country i is higher under
MFN relative to discriminatory treatment. For the same number of firms mi ¼ m;the first source country gains less from MFN relative to discriminatory treatment
than the second source country due to its larger labor supply
DW sp12 ; DW sp
1 2 DW sp2 ¼ 2
cDk
4MK þ 3k0 2 4
tm
c
� �
, 0: ð27Þ
A.J. Glass and K. Saggi290
The host country suffers when it loses its freedom to set FDI policy differentially.
Since world welfare if fixed, the second source country gains from imposing MFN
treatment whereas the first source country may gain or lose.
13.5.4. MFN versus national treatment
Removing a symmetric tax decreases the profits of host firms and increases the
profits of source firms. Labor earnings decrease in both source countries and
increase in the host country. Thus, each country has a conflict between labor
earnings and profits in overall welfare.
Let DWnsi ; Wn
i 2W si be the extent that welfare in country i is higher under
national relative to MFN treatment. For the same number of firmsmi ¼ m; the firstsource country gains less from national relative to MFN treatment than the second
source country due to its larger labor supply.13
DWns12 ; DWns
1 2 DWns2 ¼ 2
tmDk
M, 0: ð28Þ
The host country suffers when it loses its freedom to discriminate against foreign
relative to domestic firms. The second source country gains from a move from
MFN to national treatment, whereas the first source country may gain or lose.
Proposition 6. The smaller source country, which has the larger natural
incentive to conduct FDI, benefits from an investment agreement adopting MFN
treatment standards. The host country necessarily loses, and the other source
country may benefit or lose. These same effects hold for adopting national
treatment, whether start from discriminatory or MFN treatment.
13.6. ANOTHER BASIS FOR DISCRIMINATION
Our model points out that the optimal host policy differs across the two source
countries due to differences in labor supplies in the two source countries that
generate different incentives for FDI on the part of firms. An alternative rationale
for such discrimination might stem from firms from one country having a more
efficient technology of production. How do our results change under this scenario?
13 When the symmetric policy is a subsidy, this expression holds for “same as” national treatment; “no
less favorable” national treatment would permit discrimination against local investors and thus would
be equivalent to MFN treatment—UNCTAD (1999b,c).
Crowding Out and Distributional Effects of FDI Policies 291
To highlight the role of technology, assume that labor supplies and number of
firms are symmetric across countries ki ¼ k and mi ¼ m: Furthermore, let firms in
one of the source countries (say the second country) have a less efficient technology
of production: they require u $ 1 units of labor to produce one unit of output.
Equilibrium under autarky can be derived as before. Nothing changes for the
host and the first source country. However, the first order condition of firms from
the second source country is given by
p2 y2c ¼ uz2: ð29Þ
Furthermore, the labor market constraint for the second source country also needs
to be rewritten asumy2 ¼ k: ð30Þ
When FDI is allowed, the equilibrium can also be derived as before, except that
the labor constraint in the host country must be rewritten as
my0 þ a1my1 þ ua2my2 ¼ k; ð31Þ
and the labor constraint in the second source country must be rewritten as
uð12 a2Þmy2 ¼ k: ð32Þ
We can derive results similar to those derived under the basic model. The main
motivation for introducing technology differences is to examine the rationale for
discrimination. Similar results can be derived in this alternative model: firms from
the country with the smaller desire to engage in FDI face a smaller tax and the
symmetric tax is bound by the two asymmetric taxes. Not only do the various
effects discussed in this chapter continue to exist, but the consequences of non-
discrimination are analogous.
13.7. CONCLUSION
This chapter examines the cross-country repercussions of FDI policies. A tax
reduction on FDI from one country promotes FDI from the favored source country
while discouraging FDI from the other source country (crowding out effect),
lowers wages in the favored source country while raising wages in the other source
country and in the host country (distributional effect), and expands output for
multinationals from the favored source country while contracting output for
multinationals from the other source country and local firms in the host country
A.J. Glass and K. Saggi292
(strategic effect). Distributional consequences arise, both within as well as across
countries.
Our results help evaluate some of the potential distributional consequences of
adopting MFN or national treatment standards. Nondiscriminatory treatment of
multinationals forces the host country to increase the tax on firms from the source
country with the smaller desire to engage in FDI and lower the tax on firms from
the other source country. These findings help portray the positions different
interest groups (as well as different countries) can be expected to take on the issues
of MFN and national treatment.
The principles of MFN and national treatment appear in existing investment
agreements and can be expected to figure prominently during negotiations for a
potential multilateral investment agreement. Our chapter takes a useful step
forward by exploring the consequences of adopting these principles in
international investment policies. Our work focuses on distributional effects: we
do not address issues of efficiency since total output is fixed in our model by
design. Future research should, no doubt, analyze efficiency.
ACKNOWLEDGEMENTS
We thank Werner Antweiler, Rick Bond, Jim Brander, Shabtai Donnenfeld,
Eckhard Janeba, Howard Marvel, Horst Raff, Maurice Schiff, Tony Venables,
participants at the Midwest International Economics meetings, the Canadian
Economic Association meetings, the World Congress of the Econometric Society,
and the European Trade Study Group for helpful comments.
APPENDIX A
Define ki ; kiK so that an increase in ki increases labor in source country i
relative to the world. Define mi ; miM and interpret m similarly but for the
number of firms.
A.1. PROOF OF PROPOSITION 1
Increasing labor in a source country relative to the world reduces the extent of FDI
into that source country
›ani
›ki¼ 2
1
m, 0:
Crowding Out and Distributional Effects of FDI Policies 293
Increasing the number of firms in a source country relative to the world raises the
extent of FDI into that source country
›an1
›m¼
hk
nm2. 0:
A.2. PROOF OF PROPOSITION 2
The extent of FDI from the first source country decreases with its own tax
›ap
1
›t1¼
2M
m1
2 1
� �
Mck1
½ðM 2 m1Þs1 2 m2s2 þ cK�2, 0;
and increases with its rival’s tax
›ap
1
›t2¼
m2
M
m1
ck1
½ðM 2 m1Þs1 2 m2s2 þ cK�2. 0:
The wage in the host country decreases with either tax
›zp0
›t1¼ 2
m1
M, 0:
The wage in the first source country increases with its own tax
›zp1
›t1¼ 12
m1
M
� �
. 0;
and decreases with its rival’s tax
›zp1
›t2¼ 2
m2
M, 0:
Output by a firm from the host country increases with either tax
›yp0
›t1¼
1
c
m1
M. 0:
A.J. Glass and K. Saggi294
Output by a firm from the first source country decreases with its own tax
›yp1
›t1¼ 2
1
c12
m1
M
� �
, 0;
and decreases with its rival’s FDI tax
›yp1
›t2¼
1
c
m2
M. 0:
A.3. PROOF OF PROPOSITION 3
Obvious from the expression for the difference in taxes (Equation 23).
A.4. PROOF OF PROPOSITION 4
The extent of FDI from source country i is
asi ¼
c K 2M
mi
ki
� �
2 m0t
cK 2 m0t; ;i ¼ 1; 2;
and similarly for the second source country. Equilibrium wages in the host country
equal
zs0 ¼ p2
cK þ tðm1 þ m2Þ
M;
and equilibrium wages in each source country equal
zsi ¼ z
s¼ p2
cK 2 tm0
M; ;i ¼ 1; 2:
Lastly, equilibrium output of a local firm in the host country is
ys0 ¼
K
Mþ
tðm1 þ m2Þ
Mc;
Crowding Out and Distributional Effects of FDI Policies 295
while equilibrium output of a multinational firm from source country i is
ysi ¼ y
s¼
K
M2
tm0
Mc; ;i ¼ 1; 2:
The extent of FDI from either source country decreases with the tax
›as1
›t¼ 2
m0
m1
Mck1
½m0sþ cK�2, 0:
The wage in the host country decreases with the tax
›zs0
›t¼ 2
m1 þ m2
M, 0:
The wage in either source country increases with the tax
›zs1
›t¼
m0
M. 0:
Output by a firm from the host country increases with the tax
›ys0
›t¼
m1 þ m2
cM. 0:
Output by a firm from either source country decreases with the tax
›ys1
›t¼ 2
m0
cM, 0:
Welfare. Comparing symmetric FDI policy to nonintervention, the profits of host
firms fall due to higher costs and lower output
P s0 2P n
0 ¼ 2tm0 12m0
M
� �
2K
Mþ
t
c12
m0
M
� �� �
, 0:
Meanwhile, the profits of source firms fall
P si 2P n
i ¼ 2tm0mi
M22K 2
tm0
c
� �
, 0; ;i ¼ 1; 2:
A.J. Glass and K. Saggi296
Also, labor earnings increase in both source countries
ðzsi 2 z
ni Þki ¼
tm0ki
M. 0; ;i ¼ 1; 2;
and fall in the host country
ðzs0 2 z
n0Þk0 ¼ 2t 12
m0
M
� �
k0 , 0:
A.5. PROOF OF PROPOSITION 5
Recall that the symmetric tax is
t s¼
c
2
K þ k0M
B2 ðM 2 m0Þ
2K 2 k0
M 2 m0
2
664
3
775;
whereas the asymmetric FDI taxes are
ti ¼c
2
K þ k0M
B2 mi 2 mj
2ki
mi
2
664
3
775; ;i ¼ 1; 2:
Subtracting t s from ti and using the property that
k0
m0
.K
M.
k1
m1
.k2
m2
delivers the result.
A.6. PROOF OF PROPOSITION 6
Obvious from the expressions for the differences in welfare (Equations 24, 27
and 28).
Crowding Out and Distributional Effects of FDI Policies 297
REFERENCES
Brainard, S. L. (1997). An empirical assessment of the proximity-concentration trade-off
between multinational sales and trade. American Economic Review, 87, 520–544.
Brander, J. A. and Spencer, B. J. (1987). Foreign direct investment with unemployment and
endogenous taxes and tariffs. Journal of International Economics, 22, 257–279.
Bughin, J. and Vannini, S. (1995). Strategic direct investment under unionized oligopoly.
International Journal of Industrial Organization, 13, 127–145.
Das, S. P. (1981). Effects of foreign direct investment in the presence of unemployment.
Journal of International Economics, 11, 249–257.
Dixit, A. K. and Grossman, G. M. (1986). Targeted export promotion with several
oligopolistic industries. Journal of International Economics, 21, 233–249.
Glass, A. J. and Saggi, K. (1999). FDI policies under shared factor markets. Journal of
International Economics, 49, 309–332.
Haufler, A. and Wooton, I. (1999). Country size and tax competition for foreign direct
investment. Journal of Public Economics, 71, 121–139.
Hoekman, B. and Saggi, K. (2000). “Multilateral disciplines for investment-related
policies?,” in Global Governance, Regionalism, and the International Economy,
P. Guerrieri and H.-E. Scharrer (eds.), Baden-Baden: Nomos-Verlagsgesellschaft,
167–192.
Horstmann, I. J. and Markusen, J. R. (1989). Firm-specific assets and the gains from direct
foreign investment. Economica, 56, 41–48.
Horstmann, I. J. and Markusen, J. R. (1992). Endogenous market structures in international
trade (natura facit saltum). Journal of International Economics, 32, 109–129.
Janeba, E. (1996). Foreign direct investment under oligopoly: profit shifting or profit
capturing? Journal of Public Economics, 60, 423–445.
Janeba, E. (1998). Tax competition in imperfectly competitive markets. Journal of
International Economics, 44, 135–153.
Lahiri, S. and Ono, Y. (1998). Foreign direct investment, local content requirement, and
profit taxation. Economic Journal, 108, 444–457.
Markusen, J. R. (1995). The boundaries of multinational enterprise and the theory of
international trade. Journal of Economic Perspectives, 9, 169–189.
Markusen, J. R. and Venables, A. J. (1998). Multinational firms and the new trade theory.
Journal of International Economics, 46, 183–203.
Markusen, J. R. and Venables, A. J. (1999). Foreign direct investment as a catalyst for
industrial development. European Economic Review, 43, 335–356.
Motta, M. (1992). Multinational firms and the tariff-jumping argument: a game theoretic
analysis with some unconventional conclusions. European Economic Review, 36,
1557–1571.
Motta, M. and Norman, G. (1996). Does economic integration cause foreign direct
investment? International Economic Review, 37, 757–783.
Raff, H. and Srinivasan, K. (1998). Tax incentives for import-substituting foreign
investment: does signaling play a role? Journal of Public Economics, 67, 167–193.
A.J. Glass and K. Saggi298
Rowthorn, R. E. (1992). Intra-industry trade and investment under oligopoly: the role of
market size. Economic Journal, 102, 402–414.
UNCTAD (1999). Trends in International Investment Agreements: An Overview, Geneva:
United Nations Publications.
UNCTAD (1999). National Treatment, Geneva: United Nations Publications.
UNCTAD (1999). Most-Favoured-Nations Treatment, Geneva: United Nations Publi-
cations.
Crowding Out and Distributional Effects of FDI Policies 299
Author Index
Acemoglu 50
Adler 115
Aldrich 178
Anderson 214
Bacchetta 32
Bagwell 70–72, 75–77, 83, 88, 198
Bailey 13, 94, 115, 123, 129, 145
Baily 13, 123, 134, 148
Baldwin 13, 32, 35, 49, 52, 58, 73, 94
Bandyopadhyay 78–80
Bardhan 74
Beaulieu 161, 162
Becker 74–76, 82, 101, 234
Bentsen 137
Bergstrand 235, 239
Bergstrom 74
Bernheim 74, 258
Berry 77, 78
Bhagwati 6, 106, 115, 130, 198, 199
BHW 101, 102, 104, 105
Bikhchandani 101
Bilsen 57
Bird 250
Boadu 94
Bohara 4, 135, 151
Bonetti 234
Borrell 178
Boyce 61
Brady 13, 94, 115, 123, 129, 134, 145,
148
Brahmbhatt 32
Brainard 278
Brander 281
Brock 88
Bughin 281
Cassing 33, 214, 222
Chib 116, 119, 121, 133, 139, 146, 150,
151, 153, 154
Coate 6, 87, 199, 200, 208, 210, 211
Cohen 129
Conybeare 94, 122, 134, 145, 150
Cox 124, 134, 145, 147, 148
Cunningham 251
Das 281
David 101
Davidson 3, 33, 46, 54, 144
Davis 49
De Melo 32, 214
Dennis 13
Destler 73, 94
Diamond 63, 65
DiPrima 61
Dixit 106, 251, 253, 256, 258, 281,
285
Drazen 253
Dutt 22, 23, 27
Dutz 214
Dybvig 101
Elliott 57, 85
Endersby 186
Esquivel 93
Ethier 100
Falvey 33
Farrell 101
Feenstra 85, 87
Feldman 6, 220, 222
Fernandez 198, 201
Findlay 11
Freeman 54, 58
Friedman 250
Furusawa 33
Gaisford 33
Gawande 4, 78–81
Glass 8, 281
Gokcekus 5
Goldberg 78–80
Gonzalez 94
Goodman 215
Greenberg 116, 119, 121, 133, 139, 150,
151, 153
Greene 131, 135, 188, 191
Grier 183, 186
Groombridge 178
Grossman 4, 12, 69–74, 77, 83, 88,
160, 162, 198–200, 202, 203, 206,
210, 251, 255, 256, 258, 274,
281, 285
Hall 4, 163
Haltiwanger 57
Hamermesh 49, 50
Harper 178
Hartigan 161
Haufler 278
Haynes 105
Heckman 125, 135
Heckscher 5, 27, 160, 161, 171
Helpman 4, 12, 69–74, 77, 83, 88, 160,
162, 198–200, 202, 203, 206, 210,
215, 251, 255, 256, 258, 274
Hillman 81, 87, 198, 214, 217, 222
Hirshleifer 101
Hoekman 278
Holian 94
Horstmann 278
Hufbauer 57, 85
IMF 250
Irwin 13, 115, 123, 124, 136, 145, 148
Jacobson 32, 57
Janeba 278
Jansen 32
Jennings 94
Johnson 198
Kaempfer 6, 214, 215, 234
Kahane 94
Kalt 114
Kamdar 94
Karabay 3
Karp 33
Katz 101
Kau 114, 115
Kiewiet 103
Kim 33
Kinder 103
Knowles 5
Konings 57
Krehbiel 13, 28
Krishna 6, 81, 84, 198–200, 202–204,
210, 211, 215, 216
Kroszner 123, 124, 136, 145, 148
Krozner 13
Krueger 84, 214
Krugman 115
Lahiri 280
Lai 33
Lapinski 115
Layard 54, 59
Leamer 84, 239
Leger 33
Leibenstein 101
Levinsohn 78, 84
Levitt 115, 124, 129, 141, 145, 148
Levy 3, 74, 93, 96
Lewis 87
Li 33
Lindbeck 28
Linnemann 239
Ljungqvist 59
Lohmann 13
Lopez 178
Author Index302
Lowenberg 234
Ludema 6, 199, 200, 208, 210, 211
Lustig 93, 215
Mackinnon 144
Magee 5, 13, 32, 35, 49, 52, 58, 73,
88, 94, 160–165, 170
Maggi 6, 78–81, 87, 199, 209–211
Marks 178
Markusen 278
Maskus 178
Matusz 3, 33, 46, 54
Mayer 7, 11, 12, 14, 19, 20, 22, 23, 25,
27, 198, 251, 256
McCubbins 124, 134, 145, 147, 148
McGillivray 125, 149, 150
McLaren 3
Meade 215
Meyer 33
Mitra 6, 22, 23, 27, 198–200, 202–204,
210, 211, 274
Mizon 144
Mondak 103
Morris 87
Moser 198
Motta 278
Mourmouras 7, 251, 256
Munger 183, 186
Mutz 103
Neary 46
Nelson 4, 12, 92, 100, 101
Ng 214
Nollen 126, 128, 129, 142
Norman 278
North 274
Ochs 33
O’Flaherty 106
O’Halloran 13
Ohlin 5, 27, 161, 171
Olson 126, 145, 160, 167, 274
Ono 280
Ortiz 215
Pakes 78
Pastor 93
Patrick 115, 130
Paul 33
Pearce 178
Peltzman 13, 72, 73, 114, 115, 144,
145, 147, 148
Pischke 50
Poole 13, 114, 125, 131, 134, 142,
144, 145
Poyhonen 238
Putnam 13
Quinn 126, 128, 129, 142
Raff 278
Rees 49
Rich 124
Richard 144
Richardson 6, 216, 220, 224, 229
Robbins 81
Rodriguez-Clare 6, 81, 87, 199,
209–211
Rodrik 83, 198, 201
Roland-Holst 214
Rosendorff 87
Rosenthal 13, 114, 125, 131, 134, 142,
144, 145
Ross 6
Rowthorn 278
Rubin 114
Sørensen 105, 106
Saggi 8, 278, 281
Saloner 101
Sargent 59
Schelling 101
Scheve 102
Schickler 124
Schneider 129
Schwab 149
Shapiro 101
Sinclair 149, 150
Author Index 303
Slaughter 102, 107
Smith 105, 106
Snyder 115, 124, 125, 135, 168
Spatt 101
Spencer 281
Srinivasan 278
Staiger 70–72, 75–77, 79, 83, 88, 198
Steagall 94
Stigler 72, 73
Stone 105
Stratmann 168
Sweeney 214
Takacs 32
Tarr 32
Taussig 178
Ten Kate 93
Thompson 94
Thorbecke 94
Tinbergen 238
Tornell 93
Tosini 115
Tower 5, 115, 161
Trefler 33, 35
Tybout 214
Udry 74
UNCTAD 278, 291
Ursprung 81, 87
US General Accounting Office 178, 180
Vannini 281
Venables 278
Vodopivec 57
Vousden 214
Wayman 163
Weibull 28
Weingast 148
Weiss 93
Welch 101
Wellisz 11
Whinston 74, 258
Whitworth 214
Willett 274
Winters 32
Wise 93
Wolak 79
Wooton 278
WTO 84
Young 88, 214
Zinkula 94
Zupan 114
Author Index304
Subject Index
adjustment costs 3, 31–36, 48, 53, 56,
58, 59, 103
Agricultural Production, Marketing,
and Stabilization of Prices
Subcommittee 5, 189, 190, 192
agricultural protection 27
Agriculture, Nutrition and Forestry
Committee 186
American electoral college system 11
anti-dumping 79, 80, 85, 96, 98
anti-dumping policy 84, 87
assembly model 22, 23, 25, 27, 28
Bayesian analysis 148
Bayesian method 119, 144
Bayesian model 144
campaign contributions 5, 87, 159, 160,
164, 169, 177, 178, 182–186, 188,
192, 193
commitment value 252
common-agency 256, 258
common agency model 254, 271
conditional foreign aid 7
conditional grant 254, 267
conditional loan 250, 266
conditional loan rollovers 254
conditionality 7, 251, 253
congressional model 11
congressional voting 113, 115
congressional voting on trade policy 13
corporate PACs 167
cost of adjustment 32, 34
Cournot oligopolists 280
crowding out effect 279, 292
developing countries 250, 251, 261,
271, 273
discrimination 291
discriminatory liberalization 213, 215,
229
discriminatory treatment 285
distributional effect 279, 292
domestic monopoly 215, 229
economic sanctions 234
electoral college system of the United
States 24
electoral college systems 24, 26, 27
endogenous unilateralism 199, 200
equilibrium tariff 13, 25
export lobby 203, 207
factor mobility 5, 159, 161, 171
fast track authority 13
final grant 251–253, 260, 262
foreign aid 2, 7
foreign direct investment (FDI) 2, 8, 95,
96, 277–279, 292
free trade agreement (FTA) 4
free trade area (FTA) 213, 215
FTA 6, 213, 215, 216, 219, 220, 222,
225, 229
FTA formation 216, 219, 220, 222, 229
GATT 70, 77, 78, 161
GATT Uruguay Round 160, 161
GATT/WTO 209
general equilibrium model 31, 33, 34
general equilibrium trade model 36
Gibbs sampler 5, 121
government welfare functions 71
grant 7, 249, 250, 252, 253, 273
gravity model 7, 233, 235, 238–240, 243
Heckscher–Ohlin–Mayer model 27
Heckscher–Ohlin model 5, 161, 171
high-tech jobs 37
high-tech workers 39
ideology 5, 113–115, 122–125, 143,
147, 148, 150, 151
IFI 7, 250–253, 262, 271
import competing industries 32
import lobby 203, 204, 207, 208
import-competing 12, 13
import-competing sector 27, 34, 48, 58
information cascade 101, 106
informational cascade model 102, 107
intellectual property rights 95, 96
interest group 2, 7, 160, 167, 249, 251,
252, 257, 261, 272–274
international financial institutions
(IFIs) 249, 250
International Monetary Fund 2
job search 33, 34, 48, 58
labor market 35, 36, 39, 41, 47, 56, 59
labor market flexibility 53, 54, 58
labor market turnover 31, 49
labor market turnover rates 34, 35, 47
labor PACs 162, 167
labor turnover rates 3
learning 100, 101, 105
learning models 106
liberalization 31, 33–36
loan rollover 249, 252, 253, 260, 262
loans 7, 249, 250, 273
lobby 74, 160–162, 167, 209
lobbies 1, 69, 74, 75, 83, 160, 169
lobby formation 6, 202–204, 207
lobby groups 162
lobbying 4, 5, 12, 70, 73, 159–162, 174,
209, 274
lobbying (special-interest politics) 2
low-tech jobs 49
low-tech sector 37, 38, 49
low-tech workers 39
majority rule 14
majority-voting (general-interest
politics) 2
Markov Chain 116
Mayer–Heckscher–Ohlin Model 19
MCMC 121, 144
MCMC method 116
median voter 11–13, 18, 20, 23, 26, 28
median-voter model 200
Metropolis–Hasting algorithm 5
MFN 279, 285, 291, 293
MFN tariffs 96
MFN treatment 290
Monte Carlo (MCMC) method 116
most favored nation 8, 278
most-favored-nation treatment 289
multinational 8, 277, 279, 285
multiple equilibria 6
multivariate probit model 113, 115, 119,
120
MVP 116, 119, 125, 131, 137
MVP model 130, 131, 150
NAFTA 12, 91–96, 98–103, 106, 160,
161
Nash Bargaining Solution 74
Nash bargaining 209
Nash equilibrium 17
national treatment 8, 277–279, 282, 285,
288, 291, 293
national welfare 7
non-discrimination 8, 277, 278
non-tariff barriers 4, 71, 84, 95, 96,
214
North American Free Trade Agreement
(NAFTA) 4, 91
NTBs 71, 78, 79, 83
Subject Index306
oligopolistic markets 278
Omnibus bill 125, 126
Omnibus Trade and Competitiveness
Act 5, 112, 115, 150
one-time grant 252–254
organized interest groups 28
PAC 159, 162, 164–166, 168, 169, 174
PAC contribution 162, 163, 166–168
Pareto-optimal 254, 270–272
Pareto-optimality 7, 253, 254, 270, 273
payback provisions 250
political action committees 5, 159, 162
political contributions 5, 7
political economy 1, 69, 70, 88, 91,
92, 99, 105
political economy models 3
political-economy theories 2
price support loans 179
Production and Price Competitiveness
Subcommittees 186
protection 11–13, 27, 70, 71, 122, 123
protectionism 32, 122, 126
protectionist 36, 178
quota 215, 216, 222
quota-protected 6, 213–215, 229
quota rent 71, 76, 77, 216, 217, 229
reciprocal liberalization 198, 204, 206,
210
Reciprocal Trade Agreements Act
(RTAA) 13
reciprocal trade liberalization 198
reciprocated unilateralism 200
reciprocity 198
retraining 33, 42, 48, 58
rollover loans 7
Rotten Kid Theorem 74, 76
RTAA 13
sanctions 233
sanctions policy 234
search 47, 58
Senate Agricultural Production, Market-
ing, and Stabilization of Prices
Subcommittee 179
Senate Agriculture, Nutrition and Forestry
Committee 5, 179, 189, 190, 192
sequential reciprocity 199
social learning 99, 107
social welfare function 4, 69, 70, 75,
87
South African trade 6, 233, 235
specific factors model 5, 14, 27
Standard Short-run Political Economy
(SSPE) models 81
Stolper–Samuelson (S–S) theorem
160
strategic complements 6
strategic effect 279, 293
sugar 5, 177–179
sugar cane state 177, 185–187, 191
sugar industry 177, 187, 192
sugar loans 179
sugar program 180–182
Super-301 199
tariff 12, 26, 27, 36, 37, 56, 59, 71, 77,
79, 82, 83, 95, 160, 222
tariff liberalization 204
Tariff Rate Quota (TRQ) 95, 96, 180
Tobit analysis 5, 177, 192
Tobit model 191
Tobit regressions 184
Trade and Competitiveness Act 150
trade balance 96, 97
trade creation 6, 213, 215, 219, 229
trade ideology 113, 125
trade liberalization 6, 31, 32, 36, 48, 59,
102, 123, 160, 161
Trade Omnibus Act 116
trade policies 11, 12, 88, 99
trade protection 70
trade sanctions 234, 243
trade treaties 13
Subject Index 307
trading costs 7, 50
training 34, 37, 38, 40, 47, 48, 58
training costs 50
truthful equilibrium 258–260, 266–268,
274
truthful political equilibrium 256
unconditional grant decisions 264
unconditional loan decisions 263
unilateral liberalization 199, 204
unilateral tariff reduction 199, 203
unilateral trade liberalization 198, 199,
210, 211
unilateralism 6, 198
Uruguay round 12, 78, 85
US sugar 182
US sugar industry 5
US sugar program 178–180, 192
US trade policy 26, 71, 83, 86, 115
VER 76–80, 83, 84, 86, 87
voluntary export 71
voter heterogeneity 13
voting behavior 114, 122, 150
welfare 70, 74, 160, 215, 219, 251–253,
261, 262, 271
welfare function 70, 71
World Bank 2, 251
World Trade Organization (WTO) 2, 88,
278
WTO Agreements 96
Subject Index308