beyond trade in goods the role of investment and knowledge capital in applied trade policy

8/14/2019 Beyond Trade in Goods the Role of Investment and Knowledge Capital in Applied Trade Policy

1/155

Graduate School ETD Form 9

(Revised 12/07)

PURDUE UNIVERSITYGRADUATE SCHOOL

Thesis/Dissertation Acceptance

This is to certify that the thesis/dissertation prepared

By

Entitled

For the degree of

Is approved by the final examining committee:

Chair

To the best of my knowledge and as understood by the student in theResearch Integrity and

Copyright Disclaimer (Graduate School Form 20), this thesis/dissertation adheres to the provisions of

Purdue Universitys Policy on Integrity in Research and the use of copyrighted material.

Approved by Major Professor(s): ____________________________________

____________________________________

Approved by:Head of the Graduate Program Date

Csilla A. Lakatos

Beyond Trade in Goods: The Role of Investment and Knowledge Capital in AppliedTrade Policy

Doctor of Philosophy

Terrie L. Walmsley

Thomas W. Hertel

Roman M. Keeney

Marinos Tsigas

Terrie L. Walmsley

Thomas W. Hertel

Kenneth A. Foster 02/04/2011


2/155

Graduate School Form 20

(Revised 9/10)

PURDUE UNIVERSITYGRADUATE SCHOOL

Research Integrity and Copyright Disclaimer

Title of Thesis/Dissertation:

For the degree of Choose your degree

I certify that in the preparation of this thesis, I have observed the provisions ofPurdue UniversityExecutive Memorandum No. C-22,September 6, 1991,Policy on Integrity in Research.*

Further, I certify that this work is free of plagiarism and all materials appearing in this

thesis/dissertation have been properly quoted and attributed.

I certify that all copyrighted material incorporated into this thesis/dissertation is in compliance with the

United States copyright law and that I have received written permission from the copyright owners formy use of their work, which is beyond the scope of the law. I agree to indemnify and save harmless

Purdue University from any and all claims that may be asserted or that may arise from any copyright

violation.

______________________________________Printed Name and Signature of Candidate

______________________________________Date (month/day/year)

*Located at http://www.purdue.edu/policies/pages/teach_res_outreach/c_22.html

Beyond Trade in Goods: The Role of Investment and Knowledge Capital in Applied Trade Policy


Csilla A. Lakatos

02/04/2011


3/155

BEYOND TRADE IN GOODS: THE ROLE OF INVESTMENT AND

KNOWLEDGE CAPITAL IN APPLIED TRADE POLICY

A Dissertation

Submitted to the Faculty

of

Purdue University

by

Csilla Lakatos

In Partial Fulfillment of the

Requirements for the Degree

of


May 2011

Purdue University

West Lafayette, Indiana


4/155

All rights reserved

INFORMATION TO ALL USERSThe quality of this reproduction is dependent on the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscriptand there are missing pages, these will be noted. Also, if material had to be removed,

a note will indicate the deletion.

All rights reserved. This edition of the work is protected againstunauthorized copying under Title 17, United States Code.

ProQuest LLC.789 East Eisenhower Parkway

P.O. Box 1346

Ann Arbor, MI 48106 - 1346

UMI 3477691

Copyright 2011 by ProQuest LLC.

UMI Number: 3477691


5/155

ii

ACKNOWLEDGMENTS

I benefited greatly from the support of many people. To all of them, I am greatly

indebted. First of all, I would like to acknowledge my adviser, Dr. Terrie Walmsley.

I greatly appreciate her guidance and patience throughout my studies. Thank you

for always being approachable and your patience during our long-distance phone

conversations. Special thanks to Prof. Thomas W. Hertel for his suggestions and

insights regarding the policy relevance of this dissertation. In addition, I would like

to thank Prof. Hertel and Dr. Walmsley for giving me the opportunity to be a part

of the team at the Center for Global Trade Analysis: I learnt a tremendous amount

during the last four years. I thank Dr. Marinos Tsigas and Dr. Roman Keeney,

members of my dissertation committee for providing helpful suggestions. I appreciate

Dr. Tsigass assistance with the data for this research.

To David for the encouragement, support and understanding that kept me going

during all these years. Your yes, you can attitude has been a constant source of

inspiration that helped me become who I am today.

Finally, none of this would have been possible without the help of my parents.

Although more than 5,000 miles away, their support allowed me to make my dreams

come true.


6/155

iii

TABLE OF CONTENTS

Page

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

ABBREVIATIONS .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

CHAPTER 1. OVERVIEW .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

CHAPTER 2. INVESTMENT CREATION AND DIVERSION EFFECTS OFTHE ASEAN-CHINA FREE TRADE AGREEMENT .. .. .. .. .. .. .. .. .. .. .. .. 52.1 Background . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . 52.2 ASEAN-China Economic Relations .. ... .. ... ... .. ... ... .. ... ... .. ... ... . 8

2.3 Modelling Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.3.1 Design Decisions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.3.2 GDyn: a Short Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.4 Bilateralizing Investment in GDyn .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 222.4.1 GDyn-CE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.4.2 GDyn-CET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.4.2.1 The elasticity of transformation. . . . . . . . . . . . . . . . . . . . . . . 292.5 Simulation Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312.6 The Economic Impact of ACFTA .... ... .. ... ... .. ... ... .. ... ... .. ... ... . 34

2.6.1 Rates of Return and Total Investment .. .. .. .. .. .. .. .. .. .. .. .. . 352.6.2 Investment Creation and Diversion .. .. .. .. .. .. .. .. .. .. .. .. .. .. 37

2.6.3 Welfare Impacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 422.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

CHAPTER 3. KNOWLEDGE CAPITAL: A FACTOR OF PRODUCTION . . . 553.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 553.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 593.3 Description of the Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62


7/155

iv

Page

3.4 The Translog Production Function .. ... .. ... ... .. ... ... .. ... ... .. ... ... . 653.5 Results . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . .. . 68

3.5.1 R&D Versus Intangibles ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... 72

3.5.2 Econometric Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 743.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

CHAPTER 4. CROSS-RETALIATION AT THE WTO: IMPACTS OF A NODEAL IN THE US-BRAZIL COTTON DISPUTE .. .. .. .. .. .. .. .. .. .. .. .. .. .. 774.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 774.2 Dispute DS267 - US Subsidies on Upland Cotton .. .. .. .. .. .. .. .. .. .. .. 814.3 The Economics of Retaliation ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... . 874.4 Modeling Framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92

4.4.1 Quantifying Intellectual Property .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. 934.4.1.1 International accounting standards. . . . . . . . . . . . . . . . . . . 93

4.4.1.2 Royalty services in Input-Output accounting. . . . . . . . 954.4.2 Royalty Services in the GTAP Model . .. .. .. .. .. .. .. .. .. .. .. .. . 98

4.5 US-Brazil Trade Relations ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... 1004.6 Simulation Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1024.7 Impacts of a No Deal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

4.7.1 Impact on Trade Flows .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... . 1084.7.2 Impact on Consumers and Producers .. .. .. .. .. .. .. .. .. .. .. .. .. 1134.7.3 Welfare Impacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114

4.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

CHAPTER 5. SUMMARY .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

LIST OF REFERENCES .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124

APPENDICESAppendix A: Derivation of the Cross-Entropy Minimization . . . . . . . . . . . . . . . . . . 130Appendix B: List of Products ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... . 133Appendix C: Monopolistic Competition Extension in GTAP . . . . . . . . . . . . . . . . . 136

VITA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137


8/155

v

LIST OF TABLES

Table Page

2.1 Bilateral China-ASEAN FDI Inflows, 1996-2004 ($mil) . . .. . . .. . . .. . . .. . . .. . 11

2.2 Elasticities of Transformation of Investment in the Literature . . . . . . . . . . . . . . 30

2.3 Modality for the Reduction and Elimination of Tariffs . . . .. . . .. . . .. . . .. . . .. . 32

2.4 Tariffs Applied and Faced by China in 2001 (%) .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 33

2.5 Cumulative % Change in RORGE(GDyn-CE) and RORGA(GDyn-CET) . 51

2.6 Cumulative % Change in Total Investment .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 52

2.7 Cumulative Welfare Changes, 2001-2020 ($mil) .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. 53

2.8 Cumulative Changes in Equity Income, 2001-2020 ($mil) .. . . .. . . .. . . .. . . .. . 53

2.9 Cumulative Changes in Bilateral Equity Income, 2001-2020 ($mil) . . . . . . . . . 54

3.1 Composition of the Wealth Nations ($ per capita in 2000) . . . .. . . .. . . .. . . .. . 56

3.2 Overall Regression Results . ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... . 73

4.1 Budgetary Transfers to the US Cotton Sector ($mil) .. .. .. .. .. .. .. .. .. .. .. .. 83

4.2 The Composition of the Royalties Sector in the US Input-Output Table . . 97

4.3 The Evolution of Trade Flows (US$mil) .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. 103

4.4 Sectoral and Regional Aggregation .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. 104

4.5 Initial and Retaliatory Tariffs on US Exports Applied by Brazil (% AVE) 105

4.6 Volume Changes in Private Consumption ($mil) .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 115


9/155

vi

Table Page

4.7 Volume Changes in Output ($mil) .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. 116

4.8 Equivalent variation ($mil) ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... . 118


10/155

vii

LIST OF FIGURES

Figure Page

2.1 ASEAN-China Trade Flows, 1995-2008 ($mil) .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 13

2.2 ASEAN-China FDI Flows, 1995-2008 ($mil) .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 13

2.3 Cumulative % Change in RENTAL and PCGDS .. .. .. .. .. .. .. .. .. .. .. .. .. .. 38

2.4 Cumulative % Change in RORGE(CE) and RORGA(CET) . .. . . .. . . .. . . .. . 38

2.5 Cumulative % Change in Total Investment .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 39

2.6 Bilateral Ownership in 2005 ... ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... 43

2.7 Bilateral Ownership in 2010 ... ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... 43

3.1 Investment in Intangibles as % of Gross Capital Formation: 1980-2007 . . . . 57

3.2 Cost Share of Knowledge Capital .. ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... . 64

3.3 Relationship Between log(Y) and log(K) by Region-Sector Pairs . . . . . . . . . . . 64

3.4 Parameter Estimates by Sector-Region Pairs .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 70

3.5 Allen Partial Elasticities of Substitution by Sector-Region Pairs . . . . . . . . . . . 72

4.1 World Price of Cotton (nominal $cents/lb) .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 82

4.2 Major Cotton Producers and Exporters (billion bales) . . . .. . . .. . . .. . . .. . . .. . 82

4.3 Reciprocity Compensation for a WTO Inconsistent Export Subsidy . . . . . . . 91

4.4 US-Brazil Bilateral Trade Flows ($bil) .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. 103

4.5 Changes in US-Brazil Exports .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... .. ... ... . 111


11/155

viii

Figure Page

4.6 Volume Changes in Bilateral Exports ($mil) .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. . 112


12/155

ix

ABBREVIATIONS

ACFTA ASEAN-China Free Trade Agreement

AES Allen Partial Elasticity of Substitution

ASEAN Association of Southeast Asian Nations

BIT Bilateral Investment Treaty

CE Cross Entropy

CES Constant Elasticity of Sustitution

CET Constant Elasticity of Transformation

CGE Computable General Equilibrium

DSB Dispute Settlement Body

EU27 European Union 27 Member Countries

FDI Foreign Direct Investment

FTA Free Trade Agreements

GATS General Agreement on Trade and Services

GATT General Agreement on Tariffs and Trade

GDP Gross Domestic Product

GDyn Dynamic GTAP model


13/155

x

GSM Export Credit Guarantee Program

GSP Generalized System of Preferences

GTAP Global Trade Analysis Project

HIC High Income Country

HS Harmonized Commodity Description and Coding System

IMF International Monetary Fund

IPR Intellectual Property Rights

LDP Loan Defficiency Payments

LIC Low Income Country

MFN Most Favoured Nation

MIC Middle Income Country

NAICS North American Industrial Classification System

NFI Net Foreign Income

OECD Organisation for Economic Cooperation and Development

R&D Research and Development

SCM Agreement on Subsidies and Countervailing Measures

SIC Standard Industrial Classification

SNA System of National Accounts

SUR Seemingly Unrelated Regression

TRIPS Agreement on Trade Related Aspects of Intellectual Property

Rights

WTO World Trade Organization


14/155

xi

ABSTRACT

Lakatos, Csilla Ph.D., Purdue University, May 2011. Beyond Trade in Goods: TheRole of Investment and Knowledge Capital in Applied Trade Policy. Major Profes-sors: Terrie Walmsley and Thomas W. Hertel.

International trade relations have long surpassed the traditional concept of ex-

change in goods. Trade related aspects of intellectual property rights and investment

measures are among the emerging trade policy issues of the 21st century. The goal

of this dissertation is to shed light on certain aspects of the role of investment and

knowledge capital/intellectual property in applied trade policy.

Essay 1 focuses on highlighting investment creation and diversion impacts of the

preferential reduction of barriers to trade. More specifically, we focus on investment

creation and diversion effects within the framework of the free trade agreement be-

tween China and ASEAN countries in a dynamic computable general equilibrium

setting. We find clear evidence of investment diversion from the regions not signatory

of the free trade agreement, however overall investment creation impacts dominate

investment diversion effects and thus result in a welfare improvement for the world

as a whole.


15/155

xii

Essays 2 and 3 are aimed to lay the foundations for quantifying knowledge capital

and intellectual property in applied empirical analysis.

In Essay 2, knowledge capital is obtained from firm level data on intellectual prop-

erty assets and is measured by the value of copyrights, patents, licenses, trademarks

and trade names, blueprints or building designs. We provide statistical evidence that

knowledge capital is an input in production and we analyze substitution possibilities

between knowledge capital and the other factors of production. Second, this work lays

the foundation for quantifying knowledge capital in a computable general equilibrium

framework.

Finally, Essay 3 examines the role of intellectual property in the context of the

dispute settlement process at the WTO. A significant contribution of this essay lies

in the method used for quantifying trade related intellectual property. In line with

international accounting standards, we model royalty services as a separate interme-

diate industry (subject to increasing returns). We explore the economy wide impacts

of a no deal in the US-Brazil upland cotton dispute. As awarded by a WTO dis-

pute settlement panel, Brazil would have been entitled to $591 million in retaliatory

sanctions in goods sectors and $238 million in intellectual property sanctions. We

find that Brazils retaliation plan would have led to welfare gains for all countries

except the US. Most importantly however, had Brazil not been allowed to retaliate in

the form of suspension of intellectual property rights, the impact of trade retaliation

alone would have been negative for both Brazil and the US, a case of shooting oneself

in the foot to shoot at the other persons foot.


16/155

1

CHAPTER 1. OVERVIEW

International trade relations have long surpassed the traditional concept of ex-

change in goods. If the subjects covered by international trade agreements and

treaties give any indication of emerging trade policy issues, then there are few trends

that become apparent. First, in the context of multilateral GATT (General Agree-

ment on Tariffs and Trade) negotiations the first seven rounds covered only liberal-

ization of trade in goods, while the eighth round (the Uruguay Round) introduced

substantial changes to the domain of GATT by including services trade, investment,

intellectual property rights and dispute settlement. In addition, The Uruguay Round

is thus considered to have three pillars: GATS (General Agreement on Trade in

Services), TRIMS (Agreement on Trade Related Investment Measures) and TRIPS

(Agreement on Trade Related Intellectual Property Rights). In addition to the mul-

tilateral framework, an increasing number of preferential trade agreements include

provisions on services trade and trade related aspects of investment and intellectual

property.

This dissertation consists of three self-contained essays that address three distinct

topics concerning the role of investment and knowledge capital/intellectual property

in applied trade policy.


17/155

2

Essay 1 (Chapter 2) focuses on highlighting investment creation and diversion

impacts of the preferential reduction of barriers to trade. The pioneering work of

Viner (1950) first challenged the universal desirability of bilateral trade agreements

by drawing attention to the possibility of significant trade diversion impacts of such

agreements: if the trade creation effects are smaller than those of trade diversion,

the preferential agreement will be welfare reducing. Similarly to this framework we

distinguish between investment creation and diversion effects and simulate the im-

plementation of the free trade agreement between China and the ASEAN (ACFTA)

countries in a dynamic general equilibrium framework. As a first step, we adapt the

dynamic GTAP model to take account of bilateral ownership of investment. Two

versions of the model are considered: the first version is an example of applied mod-

els of investment demand, while the second is a model of investment supply. The

two versions help us determine the sensitivity of results with respect to the choice of

specification of investment behaviour. Our results show that ACFTA would boost the

economies of the liberalizing regions and increase rates of return. As a result, total

investment in both ASEAN countries and China increase. We place special emphasis

on investment creation and diversion effects of ACFTA and find clear evidence of

investment diversion effects in regions not signatory to the preferential agreement.

Nevertheless, overall investment creation impact dominate investment diversion ef-

fects resulting in a positive welfare improvement for the world as a whole.

Changing focus, Essay 2 (Chapter 3) draws attention to the importance of knowl-

edge capital in production. Despite theoretical advance (new growth theory), empir-


18/155

3

ical analysis concerning knowledge capital is scarce to non-existent. Most economic

models tend to ignore knowledge capital as a factor, including it in the residual or

referring to more comprehensive concepts such as technological change, innovation,

spillovers or research and development. Here, knowledge capital is obtained from firm

level data on intellectual property assets and is measured by the value of copyrights,

patents, licenses, trademarks and trade names, blueprints or building designs. We

provide statistical evidence that knowledge capital is an input in production and we

analyze substitution possibilities between knowledge capital and the other factors of

production. Second, this work lays the foundation for quantifying knowledge capital

in a computable general equilibrium framework.

Finally, Essay 3 (Chapter 4) explores the role of intellectual property in interna-

tional trade relations focusing on the dispute settlement process at the WTO. Dispute

DS267, US subsidies on upland cotton, was initiated in 2002 by Brazil alleging that

various provisions of the US cotton programme were in violation of WTO obligations.

After almost eight years of litigation, a WTO arbitration panel granted Brazil the

right to impose trade sanctions against the US and the possibility for cross-retaliation

in the form of suspension of intellectual property rights. The day before Brazil was to

start imposing retaliatory sanctions, the parties reached a deal. This chapter explores

the economic costs of a no deal in the US-Brazil cotton dispute with special empha-

sis on intellectual property retaliation. The framework developed here is unique in

the sense that it provides the possibility for quantifying intellectual property related

issues. As awarded by a WTO dispute settlement panel, Brazil would have been


19/155

4

entitled to $591 million in retaliatory sanctions in goods sectors and $238 million in

intellectual property and services sanctions. We find that Brazils retaliation plan

would have led to welfare gains for all countries except the US. Most importantly

however, had Brazil not been allowed to retaliate in the form of suspension of intel-

lectual property rights, the impact of trade retaliation alone would have been negative

for both Brazil and the US, a case of shooting oneself in the foot to shoot at the other

persons foot.


20/155

5

CHAPTER 2. INVESTMENT CREATION AND DIVERSION EFFECTS OF

THE ASEAN-CHINA FREE TRADE AGREEMENT

2.1 Background

At the ASEAN-China summit in November 2001, China proposed the establish-

ment of a free trade area with ASEAN countries1. The agreement (ACFTA) was

signed in November 2002 and the free trade area came into effect on 1 January 2010.

ACFTA became the worlds third largest free trade area in volume after the Eu-

ropean Union and the North American Free Trade Area. China and ASEAN had

a combined GDP of $6.6 trillion, population of 1.9 billion and total trade of $4.3

trillion in 2008. Although before the 1990s there was no official relationship between

China and ASEAN as a block, between 1995-2008 bilateral trade between China and

ASEAN increased more than tenfold. By 2009, China was ASEANs second largest

trading partner (with 11.6% of total trade), while ASEAN was Chinas forth largest

(10.1% of total trade)2. In addition, ASEAN is a key investor in China, FDI flows

1Association of Southeast Asian Nations that include Brunei, Cambodia, Indonesia, Laos, Malaysia,Myanmar, the Philippines, Singapore, Thailand and Vietnam.2Source: ASEAN Trade Statistics Database and Ministry of Commerce of China


21/155

6

reached $5.46 billions, while FDI flows from China to ASEAN amounted to $1.4

billion in 2008.

ACFTA is expected to deepen regional integration and to have significant impacts

on intra-regional trade and investment. On the one hand, ASEAN will benefit from

Chinas economic growth and investment potential through access to an expanding

and diversified market. On the other hand, the increased access to the natural resource

and raw material intensive economies of ASEAN countries will benefit China. As

a drawback, ASEAN manufacturers are expected to face higher competition from

cheaper Chinese exports on both domestic and international markets.

There are few studies that quantify the economic impacts of ACFTA. ASEAN Sec-

retariat (2001) used the comparative static GTAP model and estimated that ACFTA

will increase real GDP of ASEAN and China by 0.9% and 0.3%, respectively. The

same study found that exports from China to ASEAN will increase by 55.1%, while

exports from ASEAN to China will expand by 48% as a result of the ACFTA. Tsi-

gas and Wang (2010) modify the comparative static GTAP model to include explicit

modeling of transnational supply chains and export processing zones in China. They

found that CAFTA leads to an increase of welfare of $1.3 billion and $2.9 billion for

China and ASEAN, respectively. Jiang and McKibbin (2008) quantify the impacts

of the free trade area of the Asia-Pacific (among which the ASEAN-China FTA is

one) using a suite of CGE models such as APG-cubed (a dynamic global model),

the comparative static GTAP model and CERD (a static model for China). They


22/155

7

found that Chinas benefits increases along with the increase in the coverage of the

free trade areas.

With the exception of Jiang and McKibbin (2008) all of these studies are limited

to capturing the comparative static effects of the removal of barriers to trade. An

important drawback of comparative static analysis is that it neglects dynamic effects.

Indeed, the effect of trade policies are not immediate and a number of effects are linked

with capital accumulation over time (Baldwin, 1992). In the context of ACFTA it

becomes important to take into consideration dynamic effects especially given that

the removal of tariffs will be implemented gradually over the course of 10 years.

The objectives of this chapter are twofold.

First, we aim to highlight investment creation and diversion effects of the FTA be-

tween China and ASEAN in a dynamic general equilibrium framework. The concepts

of investment creation and diversion first defined by Kindleberger (1966) evolved

in parallel with those of trade creation and diversion (Viner, 1950). Later Bald-

win, Forslid, and Haaland (1996) described investment creation as the incentives to

increase investment within the integrating region and investment diversion as the

negative effects on investment outside the region. More specifically, discriminatory

liberalization lowers the price of capital goods and shift production to countries sig-

natory of the free trade agreement. The rental price of capital increases as industries

expand. Investment in these countries increases as a response to higher rates of re-

turn. We expect to observe significant investment creation and diversion effects as a

results of this FTA due to the fact that China and ASEAN are important recipients


23/155

8

and origins of global capital flows. Investment creation and diversion as used in this

chapter follows that of Baldwin, Forslid, and Haaland (1996).

Second, we present and compare two alternative views/models of investment which

yield different investment creation and diversion effects. As a first step, we adapt the

dynamic GTAP model to take account of bilateral ownership of investment. Two

versions of the model are considered. The first version is an example of applied

models of investment demand, while the second is a model of investment supply. The

two versions are based on different assumptions in their determination of cross-border

investment. We simulate the implementation of ACFTA and we focus on the welfare

impacts of investment creation and diversion.

2.2 ASEAN-China Economic Relations

Despite the fact that China shares a common border with three (Laos, Myanmar

and Viet Nam) of the ten ASEAN member countries, before the 1990s there was

no official relation between China and ASEAN as a group. Economic relations were

boosted starting with the signing of the ASEAN-China Framework Agreement of

Comprehensive Economic Cooperation in November 2002 with the target of creating

ACFTA in 2010. The Framework Agreement resulted in successive agreements cover-

ing different areas of economic integration: Agreement on Trade in Goods (November

2004), Agreement on Trade in Services (January 2007) and the Agreement on Invest-

ment (August 2009).


24/155

9

As of January 1, 2010 ACFTA came into effect. Under this framework, China and

six ASEAN countries (Brunei, Indonesia, Malaysia, Philippines, Thailand and Singa-

pore) eliminate tariffs on 7,000 product categories covering 90% of traded goods. The

other four ASEAN nations (Cambodia, Myanmar, Laos and Viet Nam) are expected

to join by 2015.

In 2008, ASEAN-China merchandise trade totalled $175.38 billion, 13.15 times

the value of that in 1995 ($13.32 billion). ASEANs exports to China grew from $6.2

billions in 1995 to $85.55 billions in 2008, while imports from China increased from

$7.12 billions in 1995 to $89.83 billions in 2008 (see Figure 2.1). By 2009, China was

ASEANs second largest trading partner (with 11.6% of total trade), while ASEAN

was Chinas forth largest (10.1% of total trade)3.

The composition of ASEAN-China trade is concentrated in key manufacturing

sectors. ASEANs top 5 export commodities to China in 2008 included sound and

television equipment (HS4

85), nuclear reactors and machinery (HS 84), mineral fuels

and oils(HS 27), rubber (HS 40) and animal or vegetable fats (HS 15) covering 67.5%

of total exports to China. On the other hand, the top 5 import commodities from

China were sound and television equipment (HS 85), nuclear reactors and machinery

(HS 84), iron and steel (HS 72), mineral fuels and oils (HS 27) and articles of iron and

steel (HS 73) covering 66.1% of total imports from China. We note that the structure

ASEANs imports from China and that of exports to China is very similar given that

of 3 of the top 5 commodity categories traded are identical.

3Source: ASEAN Trade Statistics Database and Ministry of Commerce of China4Harmonized Commodity Description and Coding System


25/155

10

Along with trade, bilateral investment between ASEAN and China has gradually

expanded since the signing of the Framework Agreement, although it constitutes only

a modest share of the two regions total FDI inflows. In 2008, FDI flows from China

to ASEAN countries totalled $1.49 billion - a more than eight-fold increase from $0.16

billion in 1995. Chinese FDI accounted for only 2% of total ASEAN FDI (see Figure

2.2). On the other hand, ASEAN is a net investor in China with FDI flows totalling

to $5.46 billion in 2008 (5% of total FDI inflows to China).

Table 2.1 details the evolution of bilateral FDI flows between China and ASEAN

countries. We find that in 2004 the main destinations of Chinese FDI to ASEAN were

Indonesia ($0.29 billion), Singapore ($0.21 billion) and Myanmar ($0.1 billion). On

the other hand, among ASEAN countries Singapore was by far the most significant

foreign direct investor in China ($2 billion) in 2004, followed by Malaysia ($0.35

billion) and Thailand ($0.17 billion).


26/155

Table 2.1: Bilateral China-ASEAN FDI Inflows, 1996-2004 ($mil)

China -> ASEAN

1996 1997 1998 1999 2000 2001 2002 2003 2004

Brunei 12.4 14.1 15.8 0.2 0.2 3.0

Cambodia 2.9 49.2 26.2 33.0Indonesia 8.0 -44.0 -1.2 -2.8 -1.5 -0.4 294.6Lao PDR 0.4 2.7 2.8 1.1 9.1 11.8 1.3 1.8 0.1Malaysia 6.5 23.0 3.4 1.2 -1.0 16.9 13.2 1.8 2.0Myanmar 2.2 0.4 2.6 0.5 4.8 108.1Philippines 3.1 5.8 216.4 64.9 0.1 -0.2Singapore 84.5 14.6 84.2 -27.4 -7.1 91.5 -170.9 131.7 212.6Thailand 3.9 -7.8 5.1 -2.1 7.2 -2.5 20.9 23.8 -3.8Viet Nam 3.1 28.1 1.7 7.0 21.0 24.2 9.4 1.5 85.6ASEAN 116.1 88.9 288.0 43.5 26.4 143.9 -71.9 186.6 735.0

ASEAN -> China

1996 1997 1998 1999 2000 2001 2002 2003 2004

Brunei 0.1 1.8 0.2 0.1 17.4 52.6 96.1Cambodia 7.4 5.5 2.9 2.5 1.9 9.3 13.7 12.5 20.7

Indonesia 93.5 80.0 68.8 129.2 146.9 159.6 121.6 150.1 104.5Lao PDR 0.2 0.4 1.1 3.1 1.0 5.2 0.4 4.3Malaysia 460.0 381.8 340.5 237.7 202.9 263.0 367.9 251.0 385.0Myanmar 0.6 2.7 5.1 11.0 2.3 2.3 16.8 3.5 8.8Philippines 55.5 155.6 179.3 117.3 111.1 209.4 186.0 220.0 233.2Singapore 2247.2 2606.4 3404.0 2642.5 2172.2 2143.6 2337.2 2058.4 2008.1Thailand 328.2 194.0 205.4 148.3 203.6 194.2 187.7 173.5 178.3Viet Nam 1.5 1.5 14.1 0.1 0.6 1.5 2.5 3.3 1.1ASEAN 3194.0 3428.0 4223.2 3288.8 2844.6 2984.0 3255.9 2925.4 3040.5

Source: ASEAN Statistical Yearbook and China External Economic Statistical Yearbooks


27/155

12

2.3 Modelling Framework

Traditionally, trade policy analysis has been at the core of the classic CGE exer-

cise, but with the growing importance of cross-border investment flows applied general

equilibrium models are increasingly focusing on adopting mechanisms for modeling

international investment, in general, and FDI, in particular. Incorporating interna-

tional capital mobility in CGE models requires explicit tracking of capital stocks,

ownership and wealth and the corresponding welfare effects of these.

With the importance of foreign capital flows being highlighted yet again during

the recent global financial crisis, it is clear that more attention needs to be given

to its impact. However despite this, existing trends in the literature show diverging

and often contradictory specifications of international investment. Perfect mobility of

capital across national boundaries has long been challenged in the literature. Among

the most well-known, Feldstein and Horioka (1980) proposed a savings-investment

correlation as a measure of international capital mobility and found that changes

in domestic savings are almost fully passed through into domestic investment - an

indicator of imperfect capital mobility. Further, French and Poterba (1991) somewhat

similarly to Feldstein and Horioka note the empirical regularity according to which

countries tend to allocate a significant share of their portfolio to their domestic assets

and label their discovery the home bias puzzle of investment. Under-diversification

can be explained by transaction costs, discriminatory taxes, differences in preferences

across countries. In the GDyn model, used here, an attempt is made to reconcile

the empirical findings in the literature with the theory of perfect capital mobility by


28/155

13

0

20000

40000

60000

80000

100000

1995 2000 2005

ASEAN>China ASEAN5>China

1995 2000 2005

BCLMV>China

China>ASEAN China>ASEAN5

0

2000040000

60000

80000

100000

China>BCLMV

Figure 2.1.: ASEAN-China Trade Flows, 1995-2008 ($mil)

Source: ASEAN Statistical Yearbook 2008. BCLMV refersto Brunei, Cambodia, Laos, Myanmar and Viet Nam

0

1000

2000

3000

4000

5000

1995 2000 2005

ASEAN>China ASEAN5>China

1995 2000 2005

BCLMV>China

China>ASEAN China>ASEAN5

0

1000

2000

3000

4000

5000

China>BCLMV

Figure 2.2.: ASEAN-China FDI Flows, 1995-2008 ($mil)

Source: ASEAN Statistical Yearbook 2008. BCLMV refersto Brunei, Cambodia, Laos, Myanmar and Viet Nam


29/155


30/155

15

of sectoral capital. The reason for this is that to date the dynamic GTAP model has

had a more medium to long run focus.

Second, if capital is assumed to be perfectly mobile across borders there is one

aggregate international market for capital that is cleared by a unique international

rate of return. In this approach, it is required that total supply of capital is known

a-priori. On the other hand, if capital mobility is imperfect the specification will make

use of an exogenously defined elasticity (or convergence parameter as referred to in

GDyn) that describe the responsiveness of capital flows to rate of return differentials.

Before proceeding, we first need to differentiate between mobility of physical cap-

ital and mobility of capital referring to the movement of funds for investment that

through investment in the current period become capital in the next period.

Physical capital mobility is not seen as actual movement of capital by a change

in geographic location. Instead, the two alternative theories are: mobility of physical

capital through trade6

and mobility through depreciation. Mobility of physical capital

is not the main objective of the analysis in this paper and therefore by capital mobility

we refer to the mobility of resources hereafter.

Third, in the context of a multi-period dynamic optimization process capital mo-

bility could be modeled using either a forward looking or recursive dynamic speci-

fication. In recursive dynamics, agents are assumed to be myopic and the current

period decisions are based entirely on current period variables. The forward looking

6Analogous with the theory that trade in factors is seen as a substitutes for trade in goods Mundell(1957).


31/155

16

specification assumes perfect foresight and relies heavily on (and is sensitive to) the

projected path of state variables.

After deciding on the exact combination of sectoral, regional and intertempo-

ral mobility of capital used in a model, the next step is to specify the mecha-

nism/functional form of investment allocation. Applied models can be divided into

two main categories: investment demand models and investment supply models.

Applied models of investment demand distribute investment given the amount of

savings where the specification of investment demand7 is more or less closely related

to the theoretical model of Nickell (1978). Regional demand for investment is deter-

mined in an optimization framework where the firm acts as a profit maximizer, while

equilibrium is achieved either by the adjustment of an endogenous interest rate or

adjustment of the current account. Examples of such models are for example Jung

and Thorbecke (2003), Bourguignon, Branson, and De Melo (1989) and Fargeix and

Sadoulet (1994).

Despite the fact that investment demand models are built on strong theoretical

foundations, Lemelin and Decaluwe (2007) show that the implied demand elasticity of

investment in these models is often too high and can result in unstable models. In this

sense, investment demand models exhibit high fluctuations with respect to changes

in relative profitability Bourguignon, Branson, and De Melo (1989) and possibly

allow for negative investment (disinvestment) - a feature that could generate strange

welfare results. In addition, investment demand models are calibrated on bilateral

7For a review of these models see Lemelin and Decaluwe (2007).


32/155


33/155

18

suffer from the problems described above due to the inclusion of errors in expectation

and the gradual equalization of rates of return (discussed in detail below).

For the purposes of this paper we extend the GDyn model to take account of

bilateral ownership. We consider two versions of GDyn corresponding to a specifica-

tion based on investment demand and investment supply, GDyn-CE and GDyn-CET,

respectively. GDyn-CET allocates regional investment based on the CET assump-

tion so widely used in the literature, while GDyn-CE uses an adaptive adjustment

process and cross-entropy techniques to determine investment and balance the cross-

ownership matrix.

Before turning to the detailed description of these two models, we provide an

overview of the current version of the GDyn model.

2.3.2 GDyn: a Short Review

GDyn extends the standard, comparative static version of the GTAP model (Her-

tel, 1997) by introducing international capital mobility, endogenous capital accumula-

tion and adaptive expectations theory of investment in a recursive dynamics setting.

GDyn is a real assets model, i.e. investment is associated with equity: the regional

households (shareholders) own equity in the firm equal to the value of physical capital

and earn income (dividends) corresponding to their ownership share - there are no

financial markets and no differentiation between debt and equity9. The model keeps

9In this sense, we can already assert that GDyn in its current version and after introducing thenecessary modifications is suitable for accomodating modeling of greenfield FDI.


34/155

19

track of gross ownership positions and income flows associated with them and thus

compared to the comparative static version the GTAP model is augmented to improve

the representation of balance of payments relationships.

Despite the advantages offered by perfect foresight models, the solution procedure

chosen for the GDyn is a recursive one in which investors are allowed to have errors

in their expectations, i.e. a novel adaptive expectations specification of investors

behaviour. Compared to perfect foresight models the dynamic GTAP model offers

greater empirical realism, flexibility in data specification and lower computational

complexity.

GDyn inherits the treatment of savings of the comparative static GTAP model.

As implemented by Hertel (1997), the representative household allocates regional

income that would maximize the per capita utility based on a Cobb-Douglas utility

function. Real saving is a single commodity that is defined as savings deflated by the

price of savings. The Cobb-Douglas specification keeps the budget shares constant,

implicitly assuming a constant marginal propensity to save of the household.10

Capital goods are a production sector and their supply is determined by a Leontieff

type production technology. On the other hand, capital is a value added component

and is a direct input into production of all goods (except capital goods) governed

by a CES type allocation. Capital is assumed to be perfectly mobile across sectors

determining a single rental rate across sectors that clears the market.

10Golub and McDougall (2006) develop a version of the dynamic GTAP model in which the savingrate in each region is endogenous and is a function of the ratio of wealth to income.


35/155

20

As in most recursive dynamic models, each periods equilibrium determines the

level of global savings and, implicitly, the aggregate amount of investment expenditure

available in that specific period. International capital mobility is modeled using a

disequilibrium approach that reconciles investment theory with empirical findings.

The disequilibrium approach adopted here is described by two mechanisms in the

model: first, there is a gradual convergence of the expected rate of return leading

to the equalization of expected rates of return in the long run; and second, errors

in expectations with respect to the actual rate of return are eliminated over time.

Investors are assumed to respond to expected rates of return as opposed to actual

rates of return when making investment decisions allowing for errors in expectations.

For instance, when investment in the base data is low despite high actual rates of

return it is assumed to be due to errors in expectations; investors are assumed to

behave adaptively and over time these errors are eliminated and the expected rate of

return will converge toward the observed rate of return.

The GDyn model in its current form does not make use of portfolio allocation the-

ory in determining gross ownership positions, i.e. investors reactions are based only

on (expected) rates of return and hence the GDyn model is an investment demand

driven model. Moreover, domestic households hold equity directly in domestic firms,

the lack of availability of bilateral data on foreign assets, precludes the representative

household from holding equity directly in foreign firms. This lack of bilateral data

on foreign assets and liabilities compels many CGE modelers to employ a somewhat

artificial representation of foreign investment. The GDyn model overcomes this prob-


36/155

21

lem through the adoption of a fictional entity called the global trust. The global

trust collects the saving of all the regional households and allocates this to regional

investment on their behalf. The mechanism that the GDyn model uses to determine

the composition of the cross-ownership matrix over time is cross-entropy minimiza-

tion. The choice of the cross-entropy allocation of wealth is motivated by the fact

that this type of specification is able to reproduce some of the empirical findings of

the investment literature such as the home bias of puzzle of investment (French and

Poterba, 1991).

The use of the global trust, however, could lead to the distortion of foreign asset

holdings. While each regions representative household may hold a different propor-

tion of their ownership in domestic and trust assets, all households are assumed to

hold the same distribution of foreign assets across countries (including some indirect

ownership of domestic assets). As a result of changes in the relative rates of return

across regions the representative households are expected to rebalance their portfolio

of foreign asset holdings and different regions might differ in their propensity to re-

balance their portfolios across foreign assets. The use of the global trust assumes that

all households rebalance their portfolios in the same way. However, in response to a

strong and negative change in the US rate of return China may respond by invest-

ing more in East Asia and Australia, while Americans might invest more in Europe

and the Middle East. While the average response produced by the global trust, an

increase in investment everywhere outside of the US, might be correct on average, it

is interesting to see how each countrys portfolio adjusts in response to the shock.


37/155

22

2.4 Bilateralizing Investment in GDyn

This section describes two versions of the GDyn model that account for bilateral

ownership of investment.

The data of bilateral FDI stocks that serves as a base of this modeling exercise has

been built by CEPII, France and is documented in Boumellassa, Gouel, and Laborde

(2007): contrary to other data sources, this database is fully consistent, balanced and

suitable for use in CGE work. The development of this database now allows us to

replace the global trust and calibrate the model on actual bilateral investment data.

We build two new versions of the dynamic GDyn model corresponding to two

main specifications of investment behaviour.

The first version (referred to as GDyn-CE) retains the dynamic mechanisms of

GDyn for determining regional investment but alters the cross-entropy approach used

to track bilateral ownership. This version could be mainly included in the family of

applied models of investment demand: investment demand is determined within an

optimization framework where the firm acts as a profit maximizer. Investment in each

period adds to the capital stocks, while the composition of the regional wealth matrix

is updated such as to match the change in the regional savings and investment.

The second version (referred to as GDyn-CET) is an example of the applied models

of investment supply: the capital owners goal is to allocate a given investment budget

in such a way as to maximize the present value of his net worth. Total wealth is

distributed across destinations (sectors and regions) as a function of relative rates

of return subject to the diversification constraints imposed by a constant elasticity


38/155

23

of transformation function (CET). The CET allocation determines the allocation of

regional capital stocks and investment.

2.4.1 GDyn-CE

GDyn-CE offers the advantage that it requires minimum adjustments to the

present version of GDyn and the associated database. GDyn-CE preserves the dy-

namic mechanisms and the adaptive expectations theory of the standard GDyn model,

but alters the cross-entropy approach to track bilateral ownership of capital.

The GDyn-CE model with bilateral ownership of capital does not make use port-

folio allocation theory in determining gross ownership positions, instead investors

reactions are based only on (expected) rates of return as in the standard GDyn

model. It is these mechanisms that determine regional (not bilateral) investment. The

cross-entropy minimization approach of the standard GDyn model to preserve initial

wealth-allocation between domestic and foreign (trust) ownership is now extended to

preserve bilateral allocations. Thus the cross-entropy minimization approach acts as

a quasi-portfolio diversification rule given the equalization of rates of return in the

long run.

In GDyn-CE regional investment is determined by the dynamic mechanisms re-

tained from the standard GDyn based on the interaction between actual, expected

and target rates of return. Investment further determines capital stocks and conse-

quently wealth in domestic firms. Savings on the other hand determine total wealth


39/155

24

accumulated by the household. Finally, cross-entropy determines bilateral ownership

of capital.

The use of the cross-entropy approach (combined with rates of return) employed

in GDyn-CE can be motivated with the following reasons:

in the short and medium run the model allows for differences in rates of return,

thereby stopping the concentration of investment in the region with the highest

rate of return;

the entropy allocation rule preserves the so-called home bias puzzle of in-

vestment; according to this empirical regularity countries tend to allocate a

significant share of their portfolio to their domestic assets (French and Poterba,

1991);

in order to avoid negative values for both gross foreign assets and liabilities:

no matter what the exogenous shock to income/wealth variables, cross-entropy

minimization keeps the initial shares positive during the simulation.

Cross-entropy minimization in GDyn-CE is the mechanism that keeps the com-

position of the cross-ownership matrix in GDyn close to its base year structure. The

underlying assumption is that the initial composition of the wealth matrix is the op-

timal one. Households have chosen this initial composition given their preferences in

the diversification and risk dispersion. As previously pointed out, the cross-entropy

mechanism is suitable in reproducing empirical observations such as the home-bias

of investment, portfolio diversification and for its properties in keeping original posi-

tive shares positive along the simulations.


40/155

25

Cross-entropy minimization in its most general form can be defined as the min-

imization of the degree of divergence between two partitions of a given total value

(initialsh ors and updatedshrs) subject to different constraints:

min CE=r

s

shrslog shrssh ors

In this specific case we can define:

shrs = Wrs

rW Hr

sh ors = W ors

rW H or

whereW 0rs and Wrs are the bilateral cross-ownership matrices in period t and t + 1,

respectively and W Hr and W H or are total wealth of household r in period t and

t + 1, respectively. r represents the owner, while s the location of the equity.

The cross-entropy minimization could be summed up with:

minr

s

Wrslog

WrsW ors

s.t.s

Wrs = W Hr

r

Wrs = W Fs


41/155

26

In percentage change form11 the solution to the above minimization would be the

following (the detailed derivation of the solution to the cross-entropy minimization

problem is presented in the Appendix):

wrs = r+ s (2.1)

s

Wrswrs = W Hrwhr (2.2)

r

Wrswrs = W Fswfs (2.3)

Equation 2.1 is the one determining the dynamic bilateral cross-ownership matrix in

GDyn, while Lagrangian multipliers r and s are determined by Equations 2.2 and

2.3.

Please note that considering the fact that cross-entropy minimization entails an

optimization based on shares, the system of equations will become overdetermined

and singular if solved for alln shares. Therefore we just solve for n 1 shares.

2.4.2 GDyn-CET

Many of todays well known CGE models such as MIRAGE (Bchir et al., 2002),

FTAP (Hanslow, Phamduc, and Verikios, 2000), WorldScan (Lejour, Veenendaal, and

Verweij, 2006), use the CET investment supply-type specification of Petri (1997) for

cross-border investment in general, or FDI in particular. Petri first used the extended

Armington assumption and defined cross-border investment as an isoelastic supply

11Percent change variables are presented in lower case.


42/155

27

function implying that investment decisions are differentiated by country of origin and

characterized by imperfect substitution between preferences for investing in different

countries.

Investment supply is driven by the value of total wealth to be allocated across

destinations and acts as a portfolio diversification problem: the households wealth

(or in other words the value of its portfolio) is allocated at the beginning of each

period subject to a diversification constraint.

As defined by Equation 2.4, wealth in period t equals wealth in period t1 in

addition to savings of period t.

wt= wt1+ st (2.4)

The allocation of wealth across regions is the following: wealth of regionr is allocated

by a separable nested CET function optimization in which the household maximizes

its total wealth subject to the diversification constraints.

max wr =s

rorrskrs

s.t. kr =s

1/rrs k

r+1

r

rs

r+1r

wherer0 is the elasticity of transformation12 in the rth market,rorrs is the rate of

return to investor from market r in market s, krs represents the stock of investment

12As shown by Shumway and Powell (1984) in a CET the elasticity of transformation should benegative.


43/155

28

of investor from market r in market s and rs is a preference parameter calibrated

on the initial database.

Note that wealth is valued at the price that the market pays to investors for

capital stocks, i.e. the rate of return. More correctly, wealth should be valued at the

price of capital goods (as it is defined in the standard GDyn and GDyn-CE). The

solution to the above optimization expressed in wealth share terms is shown below in

Equation 2.5:

KrsrorrsKrrorar

=wrs

wr=

rorrsrorar

1r

(2.5)

In GDyn-CE savings determine total wealth owned by the household (similarly to

GDyn-CE). Total wealth determines the bilateral ownership matrix based on relative

rates of return13 according to the CET specification. The cross-ownership matrix im-

plicitly determines wealth in domestic firms that consequently determines the change

in capital stocks and investment.

It is important to underline that Equation 2.5 determines both the allocation of

investment across regions and the change in investment in GDyn-CET, compared to

GDyn-CE where investment is determined by the elimination of errors in expectations

over time and the allocation of investment across regions is determined by the cross-

13In GDyn-CE this rate of return corresponds to the gross rate of return.


44/155

29

entropy. In addition, in GDyn-CE investors respond to expected rates of return, while

in GDyn-CET to actual rates of return.

2.4.2.1. The elasticity of transformation

The CET function specified in Equation 2.5 defines the interactions of prices (in this

case rates of return) to be characterized by a single parameter - r, parameter that

defines the sensitivity of relative investment, capital stock or wealth (depending on the

exact specification) with respect to changes in relative rates of return. Considering the

fact thatr plays a crucial role in the outcome of policy experiments, it is important

to dwell on how it is defined in the context of existing CGE models that have a CET

specification.

We have surveyed the literature with respect to econometric estimates of the

elasticity of transformation of investment with respect to changes in relative rates of

return. Our main findings are the following:

first, econometric evidence14 on developed countries shows that the implied

elasticity of FDI with respect to the after-tax rate of return is close to unity.

second, estimated coefficients of rate of return to investment have been found

to be statistically insignificant for Sub Saharan African countries/developing

countries, i.e. caeteris paribus, a higher rate of return has no significant impact

on FDI flows - Asiedu (2002).

14For a survey see Swenson (1994), Slemrod (1990).


45/155

30

Table 2.2 provides examples of the elasticity of transformation of investment used

in CGE models in the literature15.

Table 2.2: Elasticities of Transformation of Investment in the Literature

ModelDART 4FTAP 1.4Lee and van der Mensbrugghe 4WorldScan 5

Note that elasticities of transformation of investment with respect to changes in

relative rates of return reported in Table 2.2 are higher than supported by econometric

evidence. However, if r = 1 then the CET reduces to a simple Cobb-Douglas

specification with fixed shares and underestimate changes in total investment. In

addition, r 1 is likely to exhibit what could be referred to as the small shares

problem of the CET (Kuiper and van Tongeren, 2006). More specifically, a CET

supply function will tend to underestimate trade/investment creation no matter how

significant reduction in barriers to trade: if there are little or no trade/investment

flows in the initial data the impact of liberalization on these flows will be insignificant.

In GDyn-CET, we chose to set the value of elasticity of transformation of wealth

(not investment) with respect to changes in relative rates of return to r = 2. Note

that the CET function specified in GDyn-CET describes changes in bilateral owner-

ship as a function of relative rates return, as opposed to investment as a function of

relative rates of return defined in the CGE models mentioned previosly. This choice

15Based on Springer (1998), Hanslow, Phamduc, and Verikios (2000), Lee and Van der Mensbrugghe(2001), Lejour, Rojas-Romagosa, and Verweij (2008).


46/155

31

allows us to directly compare GDyn-CE and GDyn-CET and also minimizes modifi-

cations to the standard GDyn. As a consequence the elasticity of transformation of

wealth r = 2 is not directly comparable with the elasticity of transformation of in-

vestment reported in Table 2.2. In addition,r = 2 in this specification is equivalent

to r >1 in the alternative specification.16

In the next section we compare the economic impact of ACFTA obtained with

GDyn-CE and GDyn-CET.

2.5 Simulation Design

The simulations cover the period 2001-2020 comprised of eight intervals17. The

model is calibrated on the GDyn 6 database18 with 2001 as the base year. Regions

have been aggregated into 14 composite ones: China, Indonesia, Malaysia, Philip-

pines, Singapore, Thailand, Viet Nam, Rest of ASEAN, Japan, Rest of East Asia,

North America, EU 27 countries, Australia/New Zealand and Rest of the World19.

The sectoral aggregation is built around 3 sectors: agriculture, manufactures and ser-

vices. The dynamic baseline used here contains variables concerning the evolution of

16Since the specification used in GDyn-CET is on wealth and not investment, an elasticity of 2 willlead to larger changes in investment than an elasticity of 2 with respect to investment. Thus, whilean elasticity with respect to investment of 4 might give reasonable results, an elasticity of wealthwith respect to wealth could easily lead to negative investment implied by wealth changes.172001-2005, 2006-2007, 2008-2009, 2010, 2011, 2012-2013, 2014-2015, 2016-2020.

18The GDyn database is the standard GTAP database augmented with foreign income variables andGDyn specific parameters (Ianchovichina, Walmsley, and McDougall, 2010)19Abbreviated with CHN, IDN, MYS, PHL, SGP, THA, VNM, XSE, JPN, EAS, NAM, EU27, AUNand ROW, respectively.


47/155

32

some macroeconomic aggregates (GDP and population) and labor supply documented

in Walmsley (2006).

To simulate the effect of the FTA between ASEAN and China we gradually re-

duce and eliminate applied MFN tariffs rates according to the Modalities for Tariff

Reduction and Elimination for Tariff Lines Placed in the Normal Track as summa-

rized in Table 2.3 below20. The dynamic nature of the model allows us to gradually

implement the elimination of barriers to trade precisely as defined in the modalities.

Accordingly, by 2010 full tariff liberalization is achieved between ASEAN countries

and China21.

Table 2.3: Modality for the Reduction and Elimination of Tariffs

2005 2007 2009 2010

X 20% 20 12 5 015%X < 20% 15 8 5 010%X < 15% 10 8 5 0

5% < X < 10% 5 5 0 0

X 5% Standstill 0 0Source: Annex 1 of the China-ASEAN Framework Agreement

X = Applied MFN tariffs

In 2001 China applied higher tariffs on ASEANs imports (11.6%) than ASEAN

applied on Chinese imports (6.7%)22. Table 2.4 details the composition of tariffs faced

and applied by China across sector and individual ASEAN countries. The highest20Annex I of the Agreement on Trade in Goods of the Framework Agreement on Economic Cooper-ation between ASEAN and China, 29 November 2004.21Sensitive products and the special modalities defined for Viet Nam, Cambodia, Lao PDR andMyanmar are not considered here.22Ad valorem equivalent tariff rates have been calculated using bilateral trade weights.


48/155

33

Table 2.4: Tariffs Applied and Faced by China in 2001 (%)

Faced tariffs IDN MYS PHL SGP THA VNM XSEAgriculture 5.2 17.7 13.7 0.4 26.5 22.0 19.2

Manufacturing 7.4 5.1 5.4 0.0 11.0 21.0 7.1

Overall 6.6 6.7 5.9 0.0 11.1 19.9 7.9Applied tariffs IDN MYS PHL SGP THA VNM XSE

Agriculture 14.0 14.2 22.2 33.4 13.8 20.5 20.7Manufacturing 11.4 10.6 9.3 11.1 18.2 10.9 5.8

Overall 11.4 10.3 10.1 10.6 16.7 12.7 7.2

Source: GTAP v6 database

tariffs to Chinese imports are applied by Viet Nam (19.9%) and Thailand (11.1%).

On the other hand, the average protection rate applied by Singapore is close to zero.

Overall, the agricultural sector is more protected than manufacturing. For instance,

Thailand applied 26.5% on agricultural products originating from China and a lower

11% on manufactures. Agricultural imports to Malaysia from China faced 17.7%

tariffs while manufacturing products 5.1%.

China applied the highest tariffs on imports originating from Thailand (16.7%).

The remaining ASEAN countries face relatively homogeneous protection ranging from

10.1% to 11.4%. The lowest tariffs (7.2%)are applied to imports from the composite

ASEAN region XSE that include Brunei, Cambodia and Laos. As in ASEAN, in

China the agricultural sector in more protected than manufacturing with respect

to ASEANs imports. For instance, tariffs applied to agricultural products from

Singapore face a 33.4% tariff while manufacturing 11.1%.


49/155

34

2.6 The Economic Impact of ACFTA

In this section, we discuss the impact of the elimination of barriers to trade be-

tween ASEAN and China. We begin with a description of the impacts on rates of

return, capital accumulation and investment before turning to the investment creation

and diversion effects of trade liberalization. We provide an analysis of the outcome of

ACFTA with the two versions of GDyn described above. The section concludes with

the analysis of long run welfare effects of ACFTA.

Results are reported as cumulative percent changes relative to the baseline: using

this approach we are able to isolate the impact of the policy shock. The liberalization

shocks have been implemented in the first four periods corresponding to 2005, 2007,

2009 and 2010 as required by the modalities described in Table 2.3 above.

Overall, we find that the impact of the elimination of barriers to trade between

China and ASEAN countries on variables such as rate of return, investment and

bilateral ownership is small. This can be explained by two main factors: first, tariffs in

the 2001 database are low in agriculture and manufacturing and second, liberalization

of the services sector trade, or more specifically construction, would have the most


50/155

35

important impact on the price of capital goods23, but services trade liberalization is

not considered here.

2.6.1 Rates of Return and Total Investment

Figure 2.3 provides a comparative depiction of the evolution of rental price of

capital (RENTAL) and price of capital goods (PCGDS) as a result of the reduction

of barriers to trade24. We find that the rental price of capital increases in all liber-

alizing regions. By 2010 (the last year in which tariff shocks are implemented) we

find a surge in the rental price in Viet Nam (4.3%), Thailand (3.1%) and Singapore

(1.89%). Rental price of capital in China increases by a modest 0.7%. This increase

is mainly due to the increased demand for capital by industries, but in particular by

manufacturing industries, that expand as a result of trade liberalization.

The impact of liberalization on the price of capital goods is smaller. By 2010,

it increases by 1.02% in Thailand, 0.69% in Singapore and 0.21% in China. On the

other hand, it declines in Viet Nam (-0.12%) and Rest of ASEAN (-0.74%). Changes

in the price of capital goods can be explained by the price of intermediates (domestic

or imported) used as inputs in the formation of the capital goods. While the price

imported intermediates purchased by capital is declining, the increased demand for

domestic intermediates fuels the rise in the price of domestic intermediates. The share23Construction represents on average 60% of the inputs in the capital goods sector24Results have been reported for GDyn-CE as differences with respect to variables such as rentalprice, price of capital goods, rates of return between GDyn-CE and GDyn-CET are small.


51/155

36

of domestic versus imported intermediates determines the overall increase or decline

in the price of capital goods.

The combined effect of the change in rental price and that of the price of capital

goods determines the actual rate of return. A first look at Figure 2.3 indicates that we

should expect the biggest change in rate of return for Viet Nam, Thailand, Singapore

and Malaysia. These expectation are confirmed by Figure 2.4. Accordingly, trade

liberalization under ACFTA appears to have the largest impact on actual rates of

return in Viet Nam, Thailand and Malaysia. By 2010, the increase in actual rate of

return in these countries reaches 4.94%, 1.85% and 1.18%, respectively (a detailed

evolution of the rates of return is also presented in Table 2.5).

Figure 2.4 also compares the evolution of actual and expected rates of return in

GDyn-CET and GDyn-CE. It is important to make this distinction as investment

in the two models is determined by different rates of return. Thus, in GDyn-CET

total investment is a function of actual rates of return as described in Equation 2.5.

More specifically, in GDyn-CET the dynamic mechanisms are no longer active in

defining total regional investment demand, instead investment is determined by the

CET allocation. The errors in expectations merely adjust to ensure that regional

investment from both the CET and the dynamic mechanisms are equal. On the other

hand, in GDyn-CE investors react to changes in expected rate of return.

According to Figure 2.4 there are important differences between actual and ex-

pected rates of return for Viet Nam, Thailand and Malaysia. Note that changes in


52/155

37

actual rate of return are higher than in the expected rate of return, pointing to the

fact that total investment in GDyn-CET should be more volatile than in GDyn-CE.

Figure 2.5 and Table 2.6 compare the evolution of total investment for GDyn-

CE and GDyn-CET. As pointed out before, differences between total investment in

the two models will be more pronounced the bigger the difference between expected

and actual rates of return. Overall, changes in total investment in GDyn-CET are

higher than in GDyn-CE. Both models reflect the increase in total investment in

all liberalizing regions (except in Rest of ASEAN) and a decrease in the rest of the

regions.

In GDyn-CE, by 2010 we find the most significant increase in investment for Viet

Nam (16.08%), Thailand (5.7%) and Malaysia (3.2%). In GDyn-CET investment

increases by 17.96% in Viet Nam and 3.47% in Malaysia. By 2020, the differences

between the two models are more pronounced explained by the fact that in the long

run there is a convergence toward steady state25

in GDyn-CE but not in GDyn-CET.

2.6.2 Investment Creation and Diversion

In economic theory, the universal desirability of preferential trade agreements

has been first challenged in Viner (1950) by drawing attention to the possibility of

significant trade diversion impacts of such agreements. Trade diversion occurs if the

free trade area diverts trade away from a more efficient supplier non-signatory of the

25Steady state occurs if there is no change in the normal rate of growth of capital stock (KHAT).The normal rate of growth is the rate at which the capital stock can grow without affecting the rateof return.


53/155

38

2

0

2

4

AUN

2005 2010 2015 2020

CHN EAS

2005 2010 2015 2020

EU27 IDN

JPN MYS NAM PHL

2

0

2

4

ROW

2

0

2

4

2005 2010 2015 2020

SGP THA

2005 2010 2015 2020

VNM XSE

PCGDSRENTAL

Figure 2.3.: Cumulative % Change in RENTAL and PCGDS

Source: Authors simulations

1

0

1

2

3

4

5

AUN

2005 2010 2015 2020

CHN EAS

2005 2010 2015 2020

EU27 IDN

JPN MYS NAM PHL

1

0

1

2

3

4

5

ROW

1

0

1

2

3

4

5

2005 2010 2015 2020

SGP THA

2005 2010 2015 2020

VNM XSE

RORGARORGE

Figure 2.4.: Cumulative % Change in RORGE(CE) and RORGA(CET)



54/155

39

0

5

10

15

20

25

AUN

2005 2010 2015 2020

CHN EAS

2005 2010 2015 2020

EU27 IDN

JPN MYS NAM PHL

0

5

10

15

20

25

ROW

0

5

10

15

20

25

2005 2010 2015 2020

SGP THA

2005 2010 2015 2020

VNM XSE

CETCE

Figure 2.5.: Cumulative % Change in Total Investment


agreement toward a less efficient supplier signatory of the agreement. If trade creation

effects are smaller than those of trade diversion, the preferential agreement will be

welfare reducing.

Similarly to this framework we distinguish between investment creation and invest-

ment diversion effects of preferential trade agreements. The concepts of investment

creation and diversion first defined by Kindleberger (1966) evolved in parallel with

those of trade creation and diversion. As in Baldwin, Forslid, and Haaland (1996), we

refer to investment creation as being the incentive to increase investment within the

area covered by the preferential agreement and to investment diversion as the negative

impact on investment not covered by the preferential agreement. More specifically,

discriminatory liberalization lowers the price of capital goods and shift production to


55/155

40

countries signatory of the free trade agreement. The rental price of capital increases

as industries expand. Investment in these countries increases as a response to higher

rates of return. We will thus observe investment creation and diversion. In some

cases there may be pure creation effects (an increase of investment relative to total

investment in the world), while in other cases investment creation may add to total

investment in the world.

We have already discussed the evolution of total investment, but for determining

the welfare impact of investment creation and diversion it is important to further

analyze bilateral changes in investment. The variable that is a proxy to bilateral

investment in GDyn is bilateral ownership.

Figures 2.6 and 2.7 depict changes in bilateral ownership in 2005 and 2010 for

both GDyn-CE and GDyn-CET. Shades of red represent an increase in ownership

while shades of blue a decrease, the more pronounced the changes the darker the

shade. Note however, that light blue is associated with a small increase in ownership.

We start with the analysis of changes in bilateral ownership in 2005 depicted in

Figure 2.6. First, note that real savings in all liberalizing countries increase: by

2005 real savings in Thailand increase by 0.79%, in Philippines by 0.61% and in Viet

Nam by 0.46% (difference between GDyn-CE and GDyn-CET in terms of changes

in savings are very small). The two models provide comparable results, however the

magnitude of changes in terms of bilateral ownership are different. A first look at the

results from the two models shows that changes in ownership are more concentrated

in GDyn-CET.


56/155

41

In GDyn-CET investors allocate more to regions with higher relative rate of return,

diverting away from the rest of the regions. All countries increase ownership of capital

stocks in Viet Nam, Thailand and the Philippines. More specifically, by 2005 we find

the highest increase in the ownership of capital stocks of all regions in Viet Nam

ranging from 0.07% increase in the domestic ownership of Viet Nam to 0.24% in the

ownership of EU27.

In GDyn-CE, the mechanisms described in Equations 2.1, 2.2 and 2.3 determine

the bilateral allocation of ownership. For instance, in Viet Nam we found that total

savings increase by 0.46%, while total investment 0.37%. Based on Equation 2.3, all

regions would want to increase investment in Viet Nam by 0.37% including Viet Nam.

Nevertheless, we know that the share of domestic ownership in the total portfolio is

very high and this drives the increase in domestic ownership down to 0.02%. At the

same time, other regions are able to invest more in Viet Nam (Australia/New Zealand

and East Asia increases ownership in Viet Nam by 0.38%).

By 2010, depicted in Figure 2.7, differences between the two models are more

pronounced. Consequently, in GDyn-CET we find that all regions concentrate more

investment in Viet Nam as relative rates of return in this country increase. Similarly,

in GDyn-CE investors increase ownership in the countries with high expected rates

of return but in this case the constraints imposed by Equations 2.2 and 2.3 lead to

better diversification than in GDyn-CET that allocates total savings on relative rates

of return.


57/155

42

The next section discusses the welfare impacts of ACFTA with emphasis on the

welfare effects due to investment creation and diversion.

2.6.3 Welfare Impacts

Understanding where welfare benefits and losses arise can provide insight not only

into the overall impact of trade liberalization, but also into how different stakeholders

in an economy are affected. In this specific case, our goal is to isolate the welfare

gains and losses that arise due to investment creation and diversion as a result of the

ASEAN China FTA.

In comparison to the static GTAP model, national accounts in GDyn have been

extended to include i

beyond trade in goods the role of investment and knowledge capital in applied trade policy

Documents