UNITED NATIONS CONFERENCE ON TRADE AND DEVELOPMENT

POLICY ISSUES IN INTERNATIONAL TRADE AND COMMODITIES

STUDY SERIES No. 15

ASSESSING REGIONAL TRADING ARRANGEMENTS

IN THE ASIA-PACIFIC

by

John GilbertDepartment of Agricultural Economics

Washington State UniversityUnited States of America

Robert ScollayAPEC Study Centre and

Economics Department, University of AucklandNew Zealand

Bijit BoraResearch Section, Trade Analysis Branch

Division on Trade in Goods and Services, and CommoditiesUnited Nations Conference on Trade and Development

Geneva, Switzerland

UNITED NATIONS

New York and Geneva, 2001

ii

NOTE

The views expressed in this study are those of the authors and do not necessarily reflect the viewsof the United Nations.

The designations employed and the presentation of the material do not imply the expression of anyopinion whatsoever on the part of the United Nations Secretariat concerning the legal status of anycountry, territory, city or area, or of its authorities, or concerning the delimitation of its frontiers orboundaries.

Material in this publication may be freely quoted or reprinted, but acknowledgement is requested,together with a reference to the document number. A copy of the publication containing the quotationor reprint should be sent to the UNCTAD secretariat:

ChiefTrade Analysis Branch

Division on International Trade in Goods and Services, and CommoditiesUnited Nations Conference on Trade and Development

Palais des NationsCH-1211 Geneva

UNCTAD/ITCD/TAB/16

UNITED NATIONS PUBLICATION

Sales No. E.01.II.D.21

ISBN 92-1-112532-4

ISSN 1607-8291

© Copyright United Nations 2001All rights reserved

iii

ABSTRACT

Using both a gravity model to consider the natural trading bloc hypothesis, and simulation usinga CGE model to make welfare estimates, we examine the potential effect of a subset of the new RTAproposal in the APEC region. In broad terms the two approaches appear consistent in their ability toidentify RTAs that are beneficial in terms of the welfare of the proposed members. However,comparison of the two alternative approaches does not lead to support for the hypothesis that naturalblocs are less likely to be damaging to those economies that remain on the outside of the new proposals.

iv

ACKNOWLEDGEMENTS

We would like to thank seminar participants at the World Bank Seminar on Future ResearchDirections in Asia held in Bangkok and at the WTO Economic Research and Analysis Division.

v

CONTENTS

I. INTRODUCTION .................................................................................................... 1

II. METHODOLOGY.................................................................................................... 3

III. ASSESSING CURRENT ARRANGEMENTS ...................................................... 5

A. Trade creating effects of RTAs............................................................................ 9B. Pooled data and trade in services ...................................................................... 10C. Trade diverting effects of RTAs......................................................................... 12D. Building blocks, stumbling blocks and

continuously welfare enhancing RTAs................................................................ 13

IV. NEW PROPOSALS AND THE “NATURAL RTA” ............................................ 15

A. Gravity results................................................................................................... 16B. CGE simulation results and sensitivity................................................................. 18C. Contrasting the techniques................................................................................. 22

V. CONCLUDING COMMENTS.............................................................................. 25

NOTES ........................................................................................................................ 26

REFERENCES ................................................................................................................... 27

Tables

1. Estimated Gravity Equations – Total Merchandise Trade 1986 to 1998......................... 62. Estimated Gravity Equations – Manufactures Trade 1986 to 1998................................. 73. Estimated Gravity Equations – Agricultural Trade 1986 to 1998.................................... 84. Estimated Gravity Equations – Pooled Data by Sector................................................. 115. Estimated Gravity Coefficients for Proposed RTAs – Pooled Data by Sector............... 176. Estimated Welfare Effect of Proposed RTAs – Equivalent Variation............................ 207. Estimated Welfare Effect of Proposed RTAs – Equivalent Variation............................ 21

1

In the early 1990s when the Asia Pa-cific Economic Cooperation Process (APEC)was gathering momentum and the outcome ofthe Uruguay Round was uncertain the coun-tries in the Asia Pacific held steadfastly to themost favoured nation principle. As a group,they intentionally avoided discussions and ex-pansion of the existing regional trading agree-ments (RTA) in the region (Bora and Findlay,1996). Since then, it appears that the enthusi-asm with the ideals of the APEC has recentlyappeared to give way to disillusionment withthe lack of progress towards achieving theambitious objectives that were set out in Bogor.By early 2000, more than twenty regional trad-ing arrangements (RTAs) have been proposedamongst various APEC members, and the listcontinues to grow. Some, such as the free-tradearrangement between New Zealand and Sin-gapore, have already been enacted. Importantresearch questions arise from these develop-ments. First, there is a need for quantitativeresearch to examine the potential effects of theproposals. Second, there is a need to under-stand how the new proposals might help orhinder the achievement of APEC’s ultimateobjectives. This paper makes a contribution tothe former.

There are two basic approaches to theempirical assessment of RTAs. The first,known as the “gravity model” approach, usesa cross-section of bilateral trade data and at-tempts to estimate a ‘normal’ trade pattern. Iforder can be found in the deviations from thatpattern, this technique can provide useful in-formation on trade effects of RTAs (in particu-lar if the cross sections are available for sev-eral time periods). Because this approach re-quires the application of statistical techniquesto existing data, it is usually used ex-post – to

confirm the presence of trade creation/diver-sion after agreements are put in place. Frankel(1997) is a comprehensive study using this tech-nique.

For situations where analysis prior to thefact is required, the most common techniquein recent years has been simulation with a com-putable general equilibrium (CGE) model. Thisapproach is quite different. It takes cross-sec-tional data from a single base period, not onlyfor trade but also production, and consumption,and imposes a detailed theoretical structure onthe interactions between different data ele-ments. These take the form of equilibrium con-straints, and assumptions on economicbehavior. The models are put to use by impos-ing changes in the underlying data (in the caseof RTAs, removing tariffs between membereconomies), and observing how the remainingvariables adjust. Many studies of this type inthe APEC context are surveyed in Scollay andGilbert (2000).

Although quite different, both tech-niques can offer insights into areas where theother is commonly used. Hence, CGE modelscan be used to consider the effect of existingarrangements through backcasting the model,or by using a past equilibrium and projectingforward in the absence of policy changes to tryand capture what the economy in questionmight have looked like without intervention.Similarly, gravity models are often used to tryand predict the outcome of proposed agree-ments by searching for pre-existing trends thatmight be interpreted as indicating “natural”blocs. The objective of this paper is to see towhat extent the predictions from these two dis-parate techniques can be correlated in the con-text of the new Asia-Pacific proposals. The

I. INTRODUCTION

2

objective is not to advance any of these pro-posals.

The paper is organized as follows. Sec-tion II outlines the methods used in the gravitymodel simulations. Section III uses the modelto assess the current state of regional tradingarrangements, in particular APEC sub-regionalgroups. This section is intended partially as aform of benchmarking, and we discuss the evi-dence provided by this approach in terms of

the traditional features of RTAs: trade creation,trade diversion and the debate over regional-ism as path towards global free trade. SectionIV uses the model to analyze a subset of newproposals, in an attempt to see whether anyconform to the ‘natural bloc’ criteria. We thencontrast these results with those obtained byexamining the same proposed blocs in a gen-eral equilibrium framework. Concluding com-ments follow in section V.

3

To analyze the effect of regional trad-ing arrangements in the Asia-Pacific context,the gravity model approach of normalizing bi-lateral trade patterns and testing for discern-able deviations from the estimated norm isadopted. The gravity model postulates that bi-lateral trade flows are proportional to the prod-uct of the size of the two economies, and in-versely related to the distance between them.This is a model that is broadly compatible witha wide variety of underlying theoretical mod-els (in particular those emphasizing imperfectcompetition – see the discussion in Frankel,1997), and that lends itself easily to empiricalverification. The basic applied model estimatesthe bilateral trade flows as a function of theproducts of the bilateral GDPs (as a measureof size), and distance (both in log form). Let-ting i and j index the economies in the modelwe have:

where Tij is the total trade betweeneconomies i and j, DISTij is the distance meas-ure, and uij is the error term. Most applicationsexpand the basic model to provide further ex-planatory variables. The model that is utilizedin this paper is of the following well-establishedform (see Frankel, 1997; Freund, 2000):

where PCi is per-capita GDP. Note thatPCi enters the equation in two forms, as theproduct of bilateral per capita GDPs, and asthe absolute value of the difference. The former

term can be thought of as capturing the impor-tance of wealth (as opposed to size) as a deter-minant of trade, the latter can be thought of ascapturing the importance of differences be-tween economies (as emphasized in theHeckscher-Ohlin type models). By virtue ofthe double-logarithmic specification of the es-timated function, the parameter estimates onthe income and distance variables (the βk ) canbe interpreted as elasticities. Hence, β1 repre-sents, for example, the estimated proportionalchange in Tij induced by a 1 per cent change inGDPiGDPj.

The remaining variables are dummiesdesigned to capture the influence of other fac-tors on trade flows. ADJij represents the exist-ence of a common border, and RTAij the exist-ence of a regional trading arrangement (being

one if both countries i and j are members of theRTA in question). OPENij is designed to cap-ture the degree of openness of RTA members(being one if country i or country j is a memberof the RTA in question), and can be thought ofas way of isolating the effect of the RTA.1 Notethat a separate RTA and OPEN dummy for each

group under consideration is utilized, and henceRTA and OPEN can be thought of as vectors ofdummy variables representing each of the in-dividual RTAs. Because the dummy variables

II. METHODOLOGY

ijijjiij uDISTGDPGDPT +++= )ln()ln()ln( 21 ββα ji <∀ (1)

ijijijij

jijiijjiij

uOPENRTAADJ

PCPCPCPCDISTGDPGDPT

++++

−++++=

321

4321 )ln()ln()ln()ln()ln(

γγγ

ββββαji <∀ (2)

4

clearly cannot be expressed in log form, theparameter estimates (γk ) are interpreted differ-ently. Hence, for example, exp( γ1 ) –1 is theproportional increase (decrease) in trade asso-ciated with having a common border. The RTAparameters can be interpreted similarly, henceexp( γ2 ) –1 is the proportional increase (de-crease) in the propensity to trade of the RTAmembers, relative to otherwise similarly sizedand located economies in the model.

The trade data comes from the Eco-nomic Research Service (ERS) time-series datain the GTAP Version 5 database (pre-releaseversion 3). The distance data is from the WorldDistance Tables (Hengeveld, 1996), and rep-resents the direct air distance between econo-mies. GDP and per-capita GDP data is fromthe World Bank World Development indica-tors database (2000), and is measured in pur-chasing power parity (PPP) terms. Using datain PPP terms allows for the avoidance of hav-ing arbitrary temporary movements in exchangerates exert undue influence over the results.However, it should be noted that obtaining ac-curate PPP measures is difficult, which couldbe an additional source of disturbance in themodel. There are a total of 38 economies inthe dataset, and hence 703 potential observa-tions in each annual period, split by agricul-tural and manufactures trade (missing valuesare dealt with simply by dropping the observa-tion from the regression). There is also a totalof 15 periods, from 1984 to 1998 (with serv-ices trade data for only one year, 1997).

The model is estimated at selected an-nual cross-sections, and also using the com-

plete pooled dataset. Using the individualcross-sections allows a chance to observechanges in the structure of world trade overtime. Using the pooled dataset also allows tobetter estimate the influence of existing or po-tential RTAs where there are limited observa-tions in the cross-sections (for example, CER).The model is estimated not only on the totalmerchandise trade, but also the individual ag-ricultural and manufactured trade datasets, ena-bling the identification of the existence ofbroad-based sectoral differences in trade pat-terns.

It is common to estimate a gravity modelusing ordinary least squares (OLS), which willproduce unbiased and consistent estimates ofthe model parameters. However, this datasetexhibits evidence of heteroskedastic errors, asis frequently the case with cross-sectional data.In this situation the efficiency of the parameterestimates can be improved through the appli-cation of generalized least squares (GLS).Since the increased error is strongly related toeconomic size (presumably reflecting measure-ment errors), the approach of weighting eachobservation by the inverse of the squared bi-lateral products of GDP is taken. In the pooleddataset with both cross-sectional and time-se-ries elements, the additional potential problemof autocorrelation is encountered. This is dealtthrough the covariance method, specifying anadditional annual dummy variable for all yearsbut the first. This technique can also be inter-preted as controlling for the growth and infla-tion in the world economy (see Bikker, 1987).

5

In the initial evaluation of the economiceffects of currently existing regional tradingarrangements, seven regional agreements atvarying degrees of development were included.Three of these, the European Union (EU),MERCOSUR (MER) and the Andean Commu-nity (AN) are groups composed of economiesentirely outside of the area of primary interest(the Asia-Pacific). Their presence in themodeling is primarily to avoid distorting re-sults by accounting for potentially influentialRTAs as a determinant of global trade patterns.The estimated effects of these agreements alsoprovide a base by which to analyze the effectof the intra-APEC groups that are the focus ofthis research. In particular, the European Un-ion, as the deepest and oldest example of re-gional trading arrangements, provides a con-venient baseline by which to evaluate the ef-fect of other arrangements.

Of the remaining four groups, three areblocs consisting entirely of a subset of APECmembers. These are NAFTA (Canada, Mexicoand the United States), AFTA (in the datasetidentified as Indonesia, Malaysia, Philippines,Singapore, and Thailand), and CER (Australiaand New Zealand). The final group is APECitself.

The results of the first set of simulations,run on selected annual cross-sections (con-tained in tables 1 through 3). In table 1 – re-sults for total merchandise trade, the first col-umn under each annual cross-section is the es-timated relationship without the OPEN vari-ables in place. The second column fits themodel with both the RTA and OPEN variables.The first point to note is that, as in other stud-ies, the gravity model does a very good job atexplaining trade patterns, with adjusted R2

measures between 0.76 and 0.86 in all of thesimulations.

The basic gravity model variables (GDP,GDPPC, and DIST) are all highly significantin most years, and take the signs expected.Trade increases with income, but at a decreas-ing rate (the parameter on GDP ranges between0.73 and 0.86). This is consistent with otherstudies. The same pattern holds with GDPPC(ranging from 0.81 to 1.07). The negative signon the distance parameter indicate that tradediminishes as distance increases, as expected(the elasticity estimates are between –0.62 and–0.83). Again, the magnitude of the estimatesis consistent with other studies. The differencein GDPPC is the only variable that does notseem to have a strong explanatory role in themodel. It is significant in only 1986, and in allcases is small. Hence, little support is foundfor the hypothesis that differences in the abso-lute value of income are a significant explana-tory factor in overall bilateral trade patternsbetween 1986 and 1998.

The adjacency variable is significant ineach year and has the expected positive effecton trade. The estimated effects are quite sub-stantial. Sharing a common border is estimatedto increase trade by between 43 and 81 per cent(exp(0.36)-1 and exp(0.61)-1), again consist-ent with the existing literature. The estimatedcoefficients on all of these variables remainsimilar in terms of both magnitude and signifi-cance when the gravity model on manufactures(table 2) and agricultural (table 3) trade is esti-mated separately. Although it is noted that thefit is not as strong in the case of agriculture asit is in the case of manufactures and total mer-chandise trade (the adjusted R2 ranges between0.52 and 0.63).

III. ASSESSING CURRENT ARRANGEMENTS

6

Table 1. Estimated Gravity Equations – Total Merchandise Trade 1986 to 1998(3-yearly intervals)

1986 1989 1992 1995 1998

(Intercept) -21.04 *** -27.82 *** -22.24 *** -27.26 *** -22.63 *** -27.93 *** -23.34 *** -28.04 *** -22.93 *** -26.59 ***(1.16) (1.30) (1.12) (1.22) (1.07) (1.15) (1.23) (1.29) (1.09) (1.17)

GDP 0.74 *** 0.84 *** 0.73 *** 0.80 *** 0.73 *** 0.80 *** 0.77 *** 0.83 *** 0.80 *** 0.86 ***(0.02) (0.03) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02)

GDPPC 0.81 *** 1.01 *** 0.88 *** 1.03 *** 0.91 *** 1.07 *** 0.91 *** 1.06 *** 0.84 *** 0.94 ***(0.04) (0.04) (0.04) (0.04) (0.03) (0.04) (0.04) (0.04) (0.03) (0.04)

DIFF 0.09 ** 0.08 * 0.05 0.04 0.04 0.03 -0.01 0.00 0.02 0.02

(0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.02)DIST -0.76 *** -0.62 *** -0.72 *** -0.62 *** -0.77 *** -0.69 *** -0.78 *** -0.71 *** -0.83 *** -0.79 ***

(0.06) (0.06) (0.06) (0.06) (0.05) (0.05) (0.06) (0.06) (0.05) (0.05)ADJ 0.48 * 0.61 ** 0.46 * 0.57 ** 0.52 ** 0.60 *** 0.48 * 0.57 ** 0.36 * 0.43 **

(0.22) (0.20) (0.21) (0.19) (0.19) (0.18) (0.21) (0.19) (0.18) (0.16)

EU 0.10 -0.04 0.10 -0.01 0.02 -0.05 -0.20 -0.24 . -0.01 -0.07(0.17) (0.15) (0.16) (0.14) (0.14) (0.13) (0.15) (0.14) (0.13) (0.12)

NAFTA -1.01 -0.77 -0.98 -0.64 -0.84 -0.48 -0.72 -0.32 -0.54 -0.12(0.68) (0.62) (0.64) (0.58) (0.59) (0.53) (0.63) (0.57) (0.55) (0.49)

AFTA 1.11 *** 0.89 ** 1.07 *** 0.73 ** 0.73 ** 0.36 0.83 ** 0.36 1.03 *** 0.65 **(0.32) (0.29) (0.30) (0.28) (0.28) (0.26) (0.30) (0.28) (0.26) (0.24)

CER 0.42 0.58 0.51 0.70 0.52 0.90 0.47 0.86 0.57 0.82

(0.96) (0.88) (0.91) (0.83) (0.84) (0.75) (0.90) (0.81) (0.78) (0.69)MERCOSUR -0.34 -0.24 0.01 -0.03 0.66 0.74 0.98 0.87 1.33 1.30

(1.09) (1.00) (1.03) (0.93) (0.94) (0.85) (1.01) (0.91) (0.88) (0.78)ANDEAN 0.57 1.22 * 0.69 1.38 ** 1.81 *** 2.32 *** 1.76 ** 2.37 *** 1.80 *** 2.44 ***

(0.57) (0.53) (0.54) (0.49) (0.49) (0.45) (0.54) (0.48) (0.46) (0.42)APEC 1.01 *** 1.10 *** 1.11 *** 1.12 *** 1.08 *** 1.11 *** 1.13 *** 1.10 *** 1.05 *** 1.12 ***

(0.11) (0.14) (0.10) (0.13) (0.09) (0.12) (0.10) (0.13) (0.09) (0.11)

EUO 0.00 -0.07 -0.10 -0.09 0.08(0.11) (0.10) (0.09) (0.10) (0.09)

NAFTAO -1.14 *** -0.97 *** -0.85 *** -0.82 *** -0.67 ***(0.13) (0.12) (0.11) (0.11) (0.10)

AFTAO 0.41 *** 0.48 *** 0.58 *** 0.61 *** 0.61 ***

(0.10) (0.10) (0.09) (0.09) (0.08)CERO -0.63 *** -0.48 *** -0.53 *** -0.57 *** -0.22 *

(0.14) (0.13) (0.12) (0.13) (0.11)MERCOSURO -0.33 * -0.16 -0.21 0.04 0.03

(0.14) (0.13) (0.12) (0.13) (0.11)ANDEANO -0.45 *** -0.55 *** -0.18 -0.38 *** -0.33 ***

(0.11) (0.11) (0.10) (0.11) (0.10)

APECO 0.22 * 0.13 0.01 0.10 -0.02(0.10) (0.10) (0.09) (0.09) (0.08)

Observations 696 696 701 701 703 703 703 703 702 702

Adjusted R2 0.76 0.80 0.78 0.82 0.81 0.85 0.78 0.82 0.83 0.86

Standard deviations in parentheses.*** Significant at greater than 1% level, ** Significant at 1% level, * Significant at 5% level.

7

Table 2. Estimated Gravity Equations – Manufactures Trade 1986 to 1998(3-yearly intervals)

1986 1989 1992 1995 1998

(Intercept) -21.76 *** -28.88 *** -22.29 *** -27.68 *** -22.89 *** -28.61 *** -23.50 *** -28.22 *** -23.20 *** -27.05 ***(1.29) (1.45) (1.19) (1.30) (1.12) (1.18) (1.22) (1.27) (1.12) (1.19)

GDP 0.75 *** 0.85 *** 0.73 *** 0.80 *** 0.74 *** 0.81 *** 0.78 *** 0.84 *** 0.81 *** 0.87 ***(0.03) (0.03) (0.02) (0.03) (0.02) (0.02) (0.02) (0.02) (0.02) (0.02)

GDPPC 0.85 *** 1.05 *** 0.89 *** 1.06 *** 0.92 *** 1.09 *** 0.91 *** 1.05 *** 0.86 *** 0.95 ***(0.04) (0.05) (0.04) (0.05) (0.04) (0.04) (0.04) (0.04) (0.04) (0.04)

DIFF 0.10 ** 0.08 * 0.04 0.03 0.03 0.02 -0.01 0.00 0.01 0.01

(0.04) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03) (0.03)DIST -0.82 *** -0.64 *** -0.78 *** -0.64 *** -0.78 *** -0.68 *** -0.80 *** -0.71 *** -0.84 *** -0.78 ***

(0.07) (0.07) (0.06) (0.06) (0.05) (0.05) (0.06) (0.06) (0.05) (0.05)ADJ 0.41 0.60 ** 0.41 0.56 ** 0.51 * 0.62 *** 0.45 * 0.58 ** 0.36 0.46 **

(0.24) (0.23) (0.22) (0.20) (0.20) (0.18) (0.21) (0.19) (0.18) (0.17)

EU 0.12 -0.04 0.07 -0.05 0.03 -0.05 -0.23 -0.27 * -0.01 -0.06(0.18) (0.17) (0.17) (0.15) (0.15) (0.13) (0.15) (0.14) (0.14) (0.12)

NAFTA -0.98 -0.81 -0.99 -0.70 -0.88 -0.51 -0.75 -0.34 -0.56 -0.13(0.75) (0.69) (0.68) (0.62) (0.61) (0.54) (0.63) (0.56) (0.56) (0.50)

AFTA 1.36 *** 1.16 *** 0.86 ** 0.54 0.75 * 0.35 0.84 ** 0.38 1.06 *** 0.66 **(0.37) (0.35) (0.32) (0.30) (0.29) (0.26) (0.30) (0.27) (0.27) (0.24)

CER 0.44 0.70 0.46 0.80 0.49 0.99 0.48 0.93 0.57 0.88

(1.07) (0.99) (0.97) (0.88) (0.87) (0.77) (0.90) (0.80) (0.80) (0.71)MERCOSUR -0.51 -0.27 -0.14 -0.13 0.50 0.67 0.85 0.82 1.27 1.30

(1.21) (1.11) (1.09) (0.99) (0.98) (0.87) (1.01) (0.89) (0.90) (0.80)ANDEAN 0.79 1.46 * 0.75 1.46 ** 1.88 *** 2.43 *** 1.81 *** 2.46 *** 1.86 *** 2.54 ***

(0.63) (0.59) (0.57) (0.52) (0.51) (0.46) (0.53) (0.48) (0.48) (0.42)APEC 1.00 *** 1.18 *** 1.13 *** 1.20 *** 1.12 *** 1.18 *** 1.13 *** 1.15 *** 1.06 *** 1.13 ***

(0.12) (0.16) (0.11) (0.14) (0.10) (0.12) (0.10) (0.13) (0.09) (0.11)

EUO 0.03 -0.09 -0.06 -0.07 0.09(0.12) (0.11) (0.10) (0.10) (0.09)

NAFTAO -1.19 *** -0.96 *** -0.87 *** -0.85 *** -0.69 ***(0.15) (0.13) (0.11) (0.11) (0.10)

AFTAO 0.31 ** 0.47 *** 0.65 *** 0.60 *** 0.64 ***

(0.12) (0.10) (0.09) (0.09) (0.08)CERO -0.81 *** -0.67 *** -0.63 *** -0.63 *** -0.28 *

(0.16) (0.14) (0.12) (0.12) (0.11)MERCOSURO -0.53 *** -0.27 -0.28 * -0.09 -0.06

(0.16) (0.14) (0.12) (0.13) (0.11)ANDEANO -0.51 *** -0.59 *** -0.17 -0.45 *** -0.39 ***

(0.13) (0.12) (0.10) (0.11) (0.10)

APECO 0.20 0.06 0.00 0.05 -0.01(0.12) (0.10) (0.09) (0.09) (0.08)

Observations 690 690 698 698 702 702 701 701 702 702

Adjusted R2 0.73 0.77 0.77 0.81 0.81 0.85 0.78 0.83 0.82 0.86

Standard deviations in parentheses.*** Significant at greater than 1% level, ** Significant at 1% level, * Significant at 5% level.

8

Table 3. Estimated Gravity Equations – Agricultural Trade 1986 to 1998(3-yearly intervals)

1986 1989 1992 1995 1998

(Intercept) -17.10 *** -21.28 *** -19.56 *** -22.81 *** -17.70 *** -18.69 *** -18.47 *** -18.19 *** -18.91 *** -18.52 ***(1.47) (1.72) (1.42) (1.62) (1.72) (1.96) (1.73) (1.95) (1.75) (1.97)

GDP 0.64 *** 0.75 *** 0.66 *** 0.75 *** 0.64 *** 0.71 *** 0.68 *** 0.71 *** 0.70 *** 0.76 ***(0.03) (0.03) (0.03) (0.03) (0.03) (0.04) (0.03) (0.04) (0.03) (0.04)

GDPPC 0.45 *** 0.57 *** 0.50 *** 0.59 *** 0.46 *** 0.49 *** 0.44 *** 0.45 *** 0.40 *** 0.39 ***

(0.05) (0.06) (0.05) (0.06) (0.06) (0.07) (0.06) (0.07) (0.06) (0.07)DIFF 0.07 0.08 0.13 ** 0.15 *** 0.07 0.09 0.11 * 0.12 ** 0.14 ** 0.13 **

(0.04) (0.04) (0.04) (0.04) (0.05) (0.05) (0.05) (0.04) (0.04) (0.04)DIST -0.45 *** -0.60 *** -0.43 *** -0.56 *** -0.55 *** -0.76 *** -0.58 *** -0.78 *** -0.57 *** -0.84 ***

(0.08) (0.08) (0.07) (0.08) (0.08) (0.09) (0.08) (0.08) (0.08) (0.08)

ADJ 0.72 ** 0.50 0.67 ** 0.49 * 0.76 * 0.38 0.75 ** 0.42 0.89 ** 0.44(0.27) (0.26) (0.26) (0.25) (0.30) (0.29) (0.28) (0.27) (0.28) (0.27)

EU 0.07 0.05 0.25 0.22 0.45 * 0.47 * 0.58 ** 0.59 ** 0.69 ** 0.68 ***(0.21) (0.20) (0.20) (0.19) (0.22) (0.22) (0.21) (0.21) (0.21) (0.20)

NAFTA -0.56 -0.12 -0.58 -0.12 0.15 0.44 0.26 0.47 0.41 0.71(0.84) (0.80) (0.79) (0.76) (0.89) (0.86) (0.85) (0.82) (0.84) (0.80)

AFTA 0.85 * 0.52 1.08 ** 0.74 * 0.26 0.25 0.19 0.07 0.19 0.17

(0.39) (0.38) (0.38) (0.36) (0.45) (0.44) (0.43) (0.42) (0.43) (0.41)CER 1.04 0.76 1.18 0.72 1.29 0.47 1.31 0.71 1.56 0.76

(1.19) (1.14) (1.13) (1.07) (1.26) (1.22) (1.21) (1.17) (1.20) (1.14)MERCOSUR 1.01 0.75 1.26 0.95 2.44 2.07 2.52 1.88 2.42 1.86

(1.34) (1.29) (1.27) (1.21) (1.42) (1.37) (1.35) (1.32) (1.35) (1.29)ANDEAN -0.93 -0.62 0.04 0.57 0.38 0.57 0.06 0.16 0.17 0.13

(0.70) (0.68) (0.67) (0.64) (0.92) (0.90) (0.72) (0.71) (0.72) (0.69)

APEC 1.11 *** 1.05 *** 1.26 *** 1.12 *** 1.12 *** 1.00 *** 1.06 *** 1.02 *** 0.96 *** 1.00 ***(0.14) (0.19) (0.13) (0.18) (0.15) (0.22) (0.14) (0.20) (0.14) (0.20)

EUO 0.19 0.20 0.17 0.17 0.32 *(0.15) (0.14) (0.16) (0.16) (0.15)

NAFTAO -0.60 *** -0.72 *** -0.33 . -0.08 0.03

(0.17) (0.16) (0.18) (0.17) (0.16)AFTAO 0.84 *** 0.58 *** 0.06 0.14 0.23

(0.14) (0.13) (0.16) (0.15) (0.15)CERO 0.45 * 0.50 ** 0.87 *** 0.64 *** 1.08 ***

(0.19) (0.18) (0.20) (0.19) (0.19)MERCOSURO 0.64 *** 0.70 *** 0.91 *** 1.18 *** 1.29 ***

(0.19) (0.17) (0.20) (0.19) (0.19)

ANDEANO 0.37 * -0.02 0.58 ** 0.31 0.69 ***(0.16) (0.15) (0.20) (0.17) (0.17)

APECO 0.22 0.44 *** 0.42 ** 0.30 * 0.18(0.14) (0.13) (0.16) (0.15) (0.14)

Observations 650 650 666 666 628 628 641 641 639 639Adjusted R2 0.53 0.57 0.58 0.63 0.52 0.56 0.55 0.58 0.56 0.61

Standard deviations in parentheses.*** Significant at greater than 1% level, ** Significant at 1% level, * Significant at 5% level.

9

A. Trade creating effects of RTAs

An attempt can be made to ascertain theeffect of RTAs in promoting intra-regional tradeby examining the estimated coefficients on theRTA dummy variables. The effects of the non-APEC related RTAs are considered first. Forthe European Union, no evidence is found of asignificant effect on total merchandise trade,in any of the years considered. Only one mar-ginally significant effect is found when manu-factures trade is separated-out (in 1995) andthis is negative. In agricultural trade, however,significant positive effect on trade post-1992is observed (e.g. after the completion of thecommon market). Agricultural trade betweenEuropean Union members ranges between 57and 99 per cent higher than would otherwisebe predicted by the gravity equation. Moreo-ver, the bias is increasing over time. These re-sults clearly reflect the pervasive influence ofthe CAP. The introduction of openness vari-ables has no significant impact on the results.

Of the two South American agreements,statistically significant results were not foundin the case of MERCOSUR. The coefficientsare positive, and in some cases quite large, from1992 onward (the agreement was formed in1991). For the much older Andean Commu-nity, the estimated intra-regional trade bias issubstantial, and statistically highly significantfrom 1992 onward. Splitting the data by sec-tor reveals that the integration is very strong inmanufactures, but less strong (and not statisti-cally significant) in the case of agriculture.

Turning to the APEC sub-regionalagreements, in the case of NAFTA, evidenceof a significant trade-creating effect was notfound. All of the coefficients on total merchan-dise and manufactures trade are negative, al-though there appears to be an increasing trend.Controlling for openness reduces the negativetrade bias in cases, but the lack of statisticalsignificance on any of the estimates makesdrawing any conclusions difficult. In the caseof agriculture, the estimated coefficients arepositive and increasing from 1992 (when the

agreement was negotiated, it was ratified twoyears later). However, again the lack of statis-tical significance makes drawing any strongconclusions difficult. Since this is, at least inpart, related to the problem of limited observa-tions on intra-NAFTA trade in the cross-sec-tional data, the question of the effect of NAFTAis returned to in the examination of the pooleddataset.

The case of AFTA provides some moreclear-cut results. From the total merchandisetrade estimates, a positive and strongly statis-tically significant bloc effect is observed. Thiseffect remains positive, and statistically signifi-cant in all years except 1992 and 1995, oncethe general openness of these economies hasbeen taken into account (as the high and verystrongly significant openness coefficients in-dicate, the economies of ASEAN are very opento trade relative to other similarly sized econo-mies – although this may be inflated somewhatby the unique role played by Singapore). Theestimates of the bias range from 43 to a stag-gering 203 per cent (144 per cent is the highestestimate when openness is included). The biaswas clearly significant prior to the decision tomove forward with an ASEAN free trade areain 1992. From the sectoral gravity equationspresented in tables 2 and 3, most of the intra-ASEAN trade bias is in manufactures trade isobserved. While there does appear to be aslightly significant positive bias in agriculturaltrade, this declines post-1992, and is of losestatistical significance. Hence it can be con-cluded that ASEAN has (thus far) only beensuccessful in promoting manufactures trade.

CER has been in place since 1983, be-fore the sample period. However, even withinthe sample period, significant evidence of tradecreating effects have not been found. Althoughthe estimates are positive and quite large (inparticular in agriculture), none are statisticallysignificant. As in the case of NAFTA, this is aproblem of limited observations on intra-CERtrade in the cross-sections which were at-tempted to be dealt with by pooling cross-sec-tional data.

10

The final test is on the significance ofan APEC group. The coefficients in the mer-chandise trade equations were found to behighly statistically significant in all years, aswell as being consistent at just over one (im-plying that members of APEC trade with oneanother roughly 2.7 times as much as other-wise similar economies). Thus, there appearsto be a definite APEC-effect that is distinct fromthe effects of regional trading arrangementswithin APEC. The estimates do not appear tobe sensitive to the inclusion of an opennessparameter. However, it was also noted that theeffect is stable over time, despite the fact thatAPEC was not formally established until 1989.Hence, while there is evidence of an effectiveintra-regional trade bias, it does not appear thatAPEC’s formal implementation has had anyeffect on that bias, or that APEC has made anyprogress in further strengthening trade ties sinceits implementation. Essentially the same pat-tern holds once manufactures are separated and(perhaps surprisingly) agricultural trade – astrong and significant regional bias, but no evi-dence of any strengthening of trade ties overtime.

B. Pooled data and trade in services

Because of the difficulty obtaining sta-tistically significant estimates of the degree oftrade integration in the Asia-Pacific using cross-sectional data, this section considers the resultsof pooling the dataset across the years 1984–1998. Through this much larger sample size,there is a greater chance of capturing the effectof arrangements among a smaller subset of thecross-section (e.g. NAFTA and CER). The re-sults are presented in table 4, for the total mer-chandise trade category, and separated bymanufactures and agricultural trade. Also pre-sented in table 4 are the estimates from apply-ing the gravity model to services trade data of1997.

As the results in table 4 indicate, thepooling technique does help in the manner in-tended. Although it makes little difference tothe parameter estimates on the basic gravity

variables (with the exception that a statisticallysignificant, but very small effect for the differ-ence in per-capita GDP is obtained), statisti-cally significant results in the case of bothNAFTA and CER are obtained (and the Euro-pean Union in the case of agriculture). In thecase of NAFTA, there is a statistically signifi-cant negative bias in overall trade and manu-factures, and a smaller but still negative bias(but insignificant) in agriculture. Introductionon an openness control lowers the bias, andmakes it positive (but still insignificant) in thecase of agricultural trade.

As for CER, statistically significant evi-dence is found of a regional trade bias in bothmanufactures and agriculture, and overall. Thebias is strongest in agricultural trade, and be-comes more positive with the control for open-ness. Thus there is evidence to suggest thatCER has been successful in promoting mer-chandise trade between Australia and NewZealand.

In the case of services, despite the lim-ited data, the gravity model again seems to pro-vide a good fit (the adjusted R2 is 0.72 withoutopenness 0.89 with).2 The coefficients on in-come (both total and per-capita) are similar tothose estimated on merchandise trade. Thecoefficient on distance, however, while stillnegative, is significantly smaller than on mer-chandise trade (-0.07 to -0.19). This indicatessupport for the hypothesis that distance is lessimportant as an explanatory factor in servicestrade.

Turning to the estimates of the effect ofRTAs on services trade, the results contrastquite strongly in places to the effects observedfor merchandise trade. There is a significantpositive effect in the case of the European Un-ion (services trade is estimated at between 27and 43 per cent higher than otherwise similareconomies). In MERCOSUR and the AndeanCommunity, in contrast to the results on mer-chandise trade, there exists a statistically sig-nificant and strongly negative services tradebias.

11

Table 4. Estimated Gravity Equations – Pooled Data by Sector

Merchandise (1984-1998) Manufactures (1984-1998) Agriculture (1984-1998) Services (1997)

(Intercept) -21.23 *** -26.14 *** -21.49 *** -26.73 *** -17.04 *** -18.30 *** -22.20 *** -27.16 ***

(0.29) (0.31) (0.30) (0.32) (0.40) (0.46) (0.97) (0.75)GDP 0.75 *** 0.82 *** 0.76 *** 0.83 *** 0.66 *** 0.73 *** 0.71 *** 0.79 ***

(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.01)GDPPC 0.86 *** 1.01 *** 0.88 *** 1.03 *** 0.45 *** 0.49 *** 0.64 *** 0.76 ***

(0.01) (0.01) (0.01) (0.01) (0.01) (0.02) (0.03) (0.03)

DIFF 0.04 *** 0.04 *** 0.04 *** 0.04 *** 0.08 *** 0.10 *** 0.02 -0.02(0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.03) (0.02)

DIST -0.77 *** -0.68 *** -0.81 *** -0.69 *** -0.51 *** -0.69 *** -0.19 *** -0.07 *(0.01) (0.01) (0.02) (0.02) (0.02) (0.02) (0.05) (0.03)

ADJ 0.45 *** 0.56 *** 0.42 *** 0.57 *** 0.73 *** 0.45 *** -0.32 0.02

(0.05) (0.05) (0.06) (0.05) (0.07) (0.07) (0.18) (0.11)EU 0.00 -0.08 * -0.01 -0.09 * 0.32 *** 0.33 *** 0.36 ** 0.24 **

(0.04) (0.04) (0.04) (0.04) (0.05) (0.05) (0.13) (0.08)NAFTA -0.84 *** -0.50 *** -0.85 *** -0.53 *** -0.19 0.18 -0.72 -0.55

(0.17) (0.15) (0.17) (0.16) (0.22) (0.21) (0.53) (0.34)AFTA 1.00 *** 0.65 *** 0.99 *** 0.63 *** 0.55 *** 0.32 ** 1.50 *** 1.08 ***

(0.08) (0.07) (0.08) (0.08) (0.11) (0.10) (0.26) (0.17)

CER 0.52 * 0.81 *** 0.50 * 0.90 *** 1.20 *** 0.69 * -1.04 -0.45(0.23) (0.21) (0.24) (0.22) (0.31) (0.30) (0.76) (0.49)

MERCOSUR 0.39 0.42 0.28 0.41 1.67 *** 1.25 *** -3.50 *** -2.51 ***(0.26) (0.24) (0.28) (0.25) (0.35) (0.34) (0.86) (0.54)

ANDEAN 1.18 *** 1.82 *** 1.27 *** 1.93 *** -0.41 * -0.18 -1.34 ** -0.61 *(0.14) (0.13) (0.14) (0.13) (0.20) (0.19) (0.45) (0.29)

APEC 1.07 *** 1.10 *** 1.09 *** 1.16 *** 1.14 *** 1.04 *** 0.25 ** 0.09

(0.03) (0.03) (0.03) (0.04) (0.04) (0.05) (0.09) (0.08)EUO -0.02 0.00 0.18 *** 0.29 ***

(0.03) (0.03) (0.04) (0.06)NAFTAO -0.89 *** -0.91 *** -0.32 *** -0.36 ***

(0.03) (0.03) (0.05) (0.07)

AFTAO 0.54 *** 0.54 *** 0.45 *** 1.01 ***(0.03) (0.03) (0.04) (0.06)

CERO -0.50 *** -0.62 *** 0.68 *** -0.35 ***(0.03) (0.04) (0.05) (0.08)

MERCOSURO -0.16 *** -0.28 *** 0.90 *** -1.12 ***(0.03) (0.04) (0.05) (0.08)

ANDEANO -0.41 *** -0.43 *** 0.32 *** -0.49 ***

(0.03) (0.03) (0.04) (0.06)APECO 0.11 *** 0.09 *** 0.28 *** 0.08

(0.02) (0.03) (0.04) (0.06)

Observations 10506 10506 10467 10467 9613 9613 703 703

Adjusted R2 0.79 0.83 0.78 0.82 0.55 0.58 0.72 0.89

Standard deviations in parentheses.*** Significant at greater than 1% level, ** Significant at 1% level, * Significant at 5% level.

12

In the case of both NAFTA and CER,the estimated coefficients are also negative(perhaps surprising in the case of CER, sincean explicit agreement to bring services tradeunder the agreement was signed in 1988).However, as was the case for merchandise traderesults estimated on single cross-sections, theresults are statistically insignificant (thereforestrong conclusions cannot be drawn).

For AFTA, there is a strong positive(and highly significant) intra-regional servicetrade bias, again indicating that this group hasbeen particularly successful in promoting in-tra-regional trade. The APEC region as a wholeis estimated to have a small positive coefficienton services trade, but this looses significanceafter control for openness.

C. Trade diverting effects of RTAs

Until this point, the openness variablepurely as a form of controlling for the generaldegree of openness of RTA members when es-timating the effect of an RTA on intra-RTAtrade has been discussed. The openness vari-able has another use, however. Observing thelevel and changes in the degree of opennesscan give insights into the presence of trade di-version effects – reductions in the level of ex-pected trade of RTA participants with non-members.

There is interest in both the level of theopenness coefficient, and any changes in thecoefficient over time (in particular if they cor-respond to post-implementation time-periods).Once again, the relevant results are in tables 1though 3, and table 4 for the pooled data.

The extra-APEC control cases are be-gun with. In the European Union, the estimatedcoefficients on openness are small, and vary-ing in sign, and statistically insignificant, henceno conclusions can be drawn. In the case ofMERCOSUR and the Andean Community, theestimated openness coefficients are negativebut diminishing overall (becoming positive for

MERCOSUR in later years, but not significant).The coefficients on agricultural trade are gen-erally positive and increasing.

Turning to the APEC sub-regionalgroups, in the case of NAFTA, the opennesscoefficient is negative, but declining over time.The pattern holds when manufactures and ag-riculture are separated (the coefficient doesbecome marginally positive in agriculture in1998, but this result is not significant). Hence,while the NAFTA economies are not as stronglyopen to trade as other economies, strong evi-dence of trade diversion effects cannot befound. In the services sector (table 4), againthere is a negative and significant coefficient(-0.34), but the lack of a time series elementmeans it cannot be observed whether this ischanging or not.

CER exhibits a similar pattern for totalmerchandise and manufactures trade – the es-timated openness coefficients are negative andsignificant, but diminishing over time. Split-ting the data along sectoral lines reveals quitea different pattern in agricultural trade, how-ever. Here, the economies of CER are shownto be very open, and moreover their degree ofopenness is increasing over time. Hence, onceagain, little evidence is found of strong trade-diversion effects. In the case of services trade,the point estimate for 1997 is negative (-0.35).

The estimated coefficients on opennessare positive in all cases for AFTA. Moreover,the estimates increase over time on both manu-factures trade, and for overall merchandisetrade. However, while the coefficients on ag-ricultural trade are estimated to be positive, theyare decreasing over time (although they are notstatistically significant post-1992). Thus thereis some indication, though not conclusive evi-dence, of trade diversion in agricultural prod-ucts occurring in ASEAN, although theseeconomies remain relatively open to agricul-tural trade.

Given APEC’s adoption to the princi-ple of “open regionalism”, which implies a

13

commitment to remaining open to non-mem-bers, evidence of trade-diversion effects asso-ciated with the APEC group is not expected(at least not in the Vinerian sense, since MFNreform cannot lead to the transfer of tariff rev-enue that is required). The simulation resultssupport this expectation – there are no statisti-cally significant negative coefficients on open-ness, no significant change post-1989, and noclear evidence of a declining trend in the open-ness coefficients over time. This conclusionapplies to both the merchandise trade data as awhole, and when separated by agricultural andmanufactures trade. The services data indicatesthat APEC is marginally more open than aver-age, but the result is not statistically signifi-cant.

D. Building blocks, stumbling blocks andcontinuously welfare enhancing RTAs

Much ink has been spilled on the topicof whether RTAs might constitute buildingblocs toward the goal of global free trade, orare instead likely to halt progress in that direc-tion (see Bhagwati and Panagariya, 1996, fordetailed discussion of these issues). An im-portant component question is whether it ispossible to arrange for RTA configurations thateliminate or minimize harmful effects on non-members, and whether such configurations arelikely to occur in practice or can be promoted.

The theoretical literature on RTAs hasanswered the existence question in the affirma-tive. It can be shown that under certain condi-tions a preferential trading arrangement willresult in a net rise in global welfare, with nonegative consequences for non-members. Thisresult is generally attributed to Kemp and Wan(1976), although there is some debate overwhether the principle was in fact recognizedearlier (see Panagariya, 2000). For the case ofa custom’s union, an intuitive explanation is asfollows: If the net trade vector of the partnereconomies was frozen with the rest of theworld, this would ensure that non-members areunaffected by the formation of the union (in a

world of perfect competition and homogene-ous goods). Then, with the external trade vec-tor as a constraint, the joint welfare of membereconomies is maximized by equating the mar-ginal rate of transformation and the marginalrate of substitution for each pair of goods acrossall agents in the union. This is accomplishedby eliminating all intra-union trade barriers, andsetting a common external tariff(endogenously) at exactly the level that satis-fies the extra-union trade constraint.Panagariya and Krishna (1997) have recentlyproved a similar result for the case of FTAs,showing that by freezing the initial vector ofimports into each member via country-specifictariffs, welfare of non-members could be main-tained intact, and the net welfare of memberscannot fall.

These theorems state that it is possibleto find preferential trading arrangements thatmust improve global welfare, but say nothingabout the consequences of any agreement ac-tually implemented in practice. Trade econo-mists are painfully aware that the existence ofwelfare gains alone is not sufficient to guaran-tee liberalization in any form. Moreover, asPanagariya (2000) has noted, in neither casecan guarantee be made that net welfare in allmembers of the union or FTA will rise – onlythat their joint welfare will rise – indeed it willgenerally be the case that one member loseswhile the other gains. It could therefore be ar-gued that divergence of interests among poten-tial members would prevent the formation ofwelfare-enhancing RTAs.

However, it is possible that diverginginterests could lead to competition to lowerexternal protection, as the KW criterion re-quires, and thus let FTA’s (at least) become truly“building blocks” toward achieving global free-trade. This possibility is explored byRichardson (1993), whose argument is essen-tially based on recognition of the presence oftrade deflection and the associated tariff rev-enue, including the idea that FTA membersmight compete for this revenue by loweringtheir tariffs to slightly below that of their part-

14

ners. When tariffs are endogenized in thismanner, the outcome is effectively free trade.Of course, other political economy models canbe set-up in which support for free trade willdecline subsequent to RTA formation (seeGrossman and Helpman, 1995; Findlay andPanagariya, 1996; and Krishna, 1998), so thequestion of whether protection declines afterRTA formation is an empirical one.

Once again, estimated coefficients onopenness, and more importantly the changesin those coefficients over time, provide someuseful information on this debate. As discussedabove, at least in the context of APEC sub-re-gional RTAs (e.g. NAFTA, AFTA and CER),evidence is found that the openness coefficientsare generally increasing over time. This is true

both overall, and for manufactures and agri-cultural trade. Since the openness coefficientscan be taken as a broad proxy measure of thelevel of protection in these RTAs, the resultsappear to lend weight to the hypothesis thatprotection levels in RTAs within APEC aredeclining, and that RTAs have not hinderedprogress toward more openness to trade in gen-eral. However, caution must be exercised soas to not conclude that the formation of RTAshas itself has promoted the openness. Althoughthe results are not inconsistent with this hypoth-esis, the data cannot render information on cau-sality. It may equally be speculated that thesuccess of negotiations under the WTO, or theinfluence of APEC, is responsible. Moreover,as always, the counter-factual is not observed.

15

There are two empirical methods com-monly applied to the analysis of RTAs. Thefirst is simulation with a gravity model, as fol-lowed in this paper. This technique is gener-ally applied ex-post, e.g. in the search for ef-fects of RTAs after they have been imple-mented. The other technique, counter-factualsimulation with partial or (more commonly inrecent years) general equilibrium trade mod-els, is frequently used to analyze the implica-tions of proposed RTAs. Scollay and Gilbert(2001) have recently used simulation tech-niques to consider the question of whether theremight be conceivable paths toward APEC goalsthrough the new Asia-Pacific RTA proposals(they identify and discuss more than twentysuch arrangements at various stages of devel-opment). Their analytical approach centers onwhether strong welfare incentives exist at eachstage of a conceivable end-point (e.g. an East-Asian bloc, or APEC liberalization) that mightlead to expansion of prior agreements. Of keyconcern in their analysis is identifying agree-ments that have strong trade diversion effects,and those that leave open the potential for wel-fare gains through further expansion.

An interesting question is whether ex-isting trade patterns provide any clues as towhen this will likely be the case, and thuswhether gravity model simulations can provideany useful analytical input. One possibility liesin extending the notion of “natural” tradingblocs. Krugman (1991) has suggested that trad-ing blocs comprised of economies that are inclose geographical proximity are unlikely toresult in significant trade diversion effects. Ofcourse, this position has been criticized on thegrounds that distance primarily affects trans-portation costs, at that these are in principle nodifferent from any other source of comparative

advantage (Bhagwati and Panagariya, 1996,among others). Evidently, this is again a mat-ter for empirical verification on a case-by-casebasis.

Gravity models contribute to the debateby using an alternative definition of “natural”.It may be speculated that in cases where strongand increasing trade integration is observedbetween economies for which there is no for-mal RTA in place, having already controlledfor distance/adjacency, there is some sort of“natural” trading bloc phenomenon being re-vealed (the results for APEC above might beinterpreted in this light). If it is held that tradeintegration in the absence of a formal agree-ment can be a useful indicator of potentiallywelfare-improving RTAs, then gravitysimulations can be useful. Hence, in this sec-tion a final set of simulations with the gravitymodel is performed, testing to see whether anyof the recently proposed agreements are in anysense “natural”, and whether this provides use-ful information on their likely effect, relativeto other potential methods.

To provide the basis of comparison, theresults of counter-factual simulations usingCGE methods are incorporated. CGE modelsare in essence numerical models based on gen-eral equilibrium theory, which are implementedin the form of computer programmes. Theyhave a number of useful features as they aremulti-sectoral and in many cases multi-re-gional, and the behavior of economic agents ismodeled explicitly through utility and profitmaximizing assumptions. In addition, they dif-fer from other multi-sector tools of analysis inthat economy-wide constraints are rigorouslyenforced. Distortions like trade barriers in aneconomic system will often have second-best

IV. NEW PROPOSALS AND THE “NATURAL RTA”

16

repercussions far beyond the sector in whichthey occur. Where the distortions are wide-ranging, general equilibrium techniques areeffective in capturing the relevant feedback andflow-through effects. The price paid, is in ad-ditional complexity of the model, and the in-ability to easily use statistical verification tech-niques.

The model utilized in this paper is thestandard GTAP (Global Trade Analysis Project)model, a publicly available model the basicstructure of which is documented in Hertel(1997). The model formulation is a standard,multi-region CGE, which assumes perfectlycompetitive markets and constant returns toscale technology. The major departure of themodel from those of standard trade theory isthe assumption of product differentiation bynational origin, controlled by a set ofArmington (1969) substitution elasticities.3

This serves the dual purpose of allowing two-way trade in each product category, and avoid-ing extreme production and trade responses.

The model is closed by assuming fullemployment of all factors of production, andthat all returns to these factors accrue to house-holds in the region in which they are employed.Final demand in each region is governed by asingle representative household, which allo-cates regional income across household expen-ditures, government spending and savings us-ing a Cobb-Douglas function.

Because CGE models attempt to cap-ture the features of real world economies, theyincorporate data on the structure of productionand trade in the economy under consideration.In general the starting point for multiregionalmodels will be a set of national input-outputand a set of trade matrices. The simulations inthis paper use the latest data available – thepre-release of the GTAP5 database, a globalgeneral equilibrium dataset has a base year of1997.4 Because part of the objective is to com-pare the two techniques, consistent data is re-quired, hence the database has been aggregatedto match the gravity sample exactly, with 33

unique regions plus a ROW aggregate, andthree sectors (agriculture, manufactures, andservices).

The following nine arrangements areconsidered (a subset of those considered inScollay and Gilbert, 2001):

• Singapore-Japan (SJ)• Singapore-United States (SUS)• Japan-Canada (JC)• Republic of Korea-Mexico (KM)• FTAA• Japan-Republic of Korea (JK)• Japan-Republic of Korea-China (JKC)• ASEAN plus Japan-Republic of Korea-

China (A3)• ASEAN plus Japan-Republic of Korea-

China plus CER (A3C)

A. Gravity results

The significance of these groupingswithin the gravity model using the same dummyvariable techniques used for existing arrange-ments above are estimated, using the pooleddata for merchandise trade (and agriculture andmanufactures separately). Because many of thearrangements involve only one observation (be-ing bilateral FTAs), this gives the greatestchance of obtaining statistically significant re-sults. The dummies for the existing RTAs areleft in place in the regressions. However, be-cause some of the arrangements listed aboveare closely related to others, each is tested sepa-rately rather than at once (thus avoiding prob-lems of collinearity in the regressors).

The results are presented in table 5 (onlythe dummy coefficients are displayed). Thegravity results for the new bilateral proposalsare mixed. The two arrangements involvingSingapore have positive and strongly signifi-cant coefficients on manufactures trade, but notagricultural trade. Thus trade ties between Sin-gapore and Japan and Singapore and the UnitedStates appear to be very strong. The coeffi-cients decline sharply (but remain positive and

17

Table 5. Estimated Gravity Coefficients for Proposed RTAs – Pooled Data by Sector1

Merchandise (1984-1998) Manufactures (1984-1998) Agriculture (1984-1998) Services (1997)

SJ 1.51 *** 0.71 ** 1.54 *** 0.70 ** 0.19 0.23 0.57 -0.44

(0.27) (0.25) (0.27) (0.25) (0.36) (0.35) (0.82) (0.52)SUS 1.91 *** 1.44 *** 1.95 *** 1.45 *** 0.75 0.60 0.98 0.04

(0.28) (0.25) (0.28) (0.26) (0.38) (0.36) (0.85) (0.52)JC -0.12 0.08 -0.20 -0.03 1.51 *** 1.90 *** -0.81 -0.65

(0.30) (0.28) (0.30) (0.28) (0.40) (0.39) (0.90) (0.57)

KM -0.37 -0.10 -0.28 -0.04 -1.95 *** -1.32 *** -0.28 0.12(0.28) (0.26) (0.28) (0.27) (0.37) (0.36) (0.85) (0.54)

FTAA 0.02 1.14 *** -0.00 1.20 *** 0.18 * 0.45 *** -1.32 *** -0.07(0.05) (0.06) (0.05) (0.06) (0.07) (0.09) (0.15) (0.13)

JK -0.36 -0.57 * -0.42 -0.64 * 0.17 0.25 -0.64 -0.43

(0.29) (0.27) (0.30) (0.28) (0.40) (0.39) (0.90) (0.57)JKC 0.40 ** 0.34 ** 0.40 ** 0.32 ** 0.10 0.32 -0.16 0.17

(0.12) (0.11) (0.12) (0.12) (0.17) (0.17) (0.37) (0.23)A3 1.15 *** 0.78 *** 1.16 *** 0.77 *** 0.52 *** 0.64 *** 0.85 *** 0.33 *

(0.06) (0.06) (0.06) (0.06) (0.09) (0.09) (0.20) (0.13)A3C 1.07 *** 0.97 *** 1.04 *** 0.99 *** 1.14 *** 0.95 *** 0.55 *** 0.29 *

(0.05) (0.06) (0.05) (0.06) (0.07) (0.08) (0.17) (0.12)

SJO 0.75 *** 0.71 *** 0.68 *** 0.32 ***(0.04) (0.04) (0.05) (0.08)

SUSO 1.04 *** 1.00 *** 1.27 *** 0.59 ***(0.04) (0.04) (0.05) (0.07)

JCO 0.00 -0.02 0.18 *** -0.16(0.04) (0.04) (0.05) (0.08)

KMO -0.02 0.02 -0.72 0.11

(0.04) (0.04) (0.05) *** (0.07)FTAAO -0.11 * -0.13 ** 0.67 *** -0.10

(0.05) (0.05) (0.07) (0.10)JKO 0.43 *** 0.43 *** 0.10 0.04

(0.04) (0.04) (0.06) (0.08)

JKCO 0.57 *** 0.60 *** 0.07 0.36 ***(0.03) (0.03) (0.05) (0.07)

A3O 0.49 *** 0.54 *** -0.10 0.35 ***(0.04) (0.04) (0.05) (0.07)

A3CO 0.46 *** 0.50 *** -0.18 ** 0.40 ***(0.04) (0.04) (0.05) (0.08)

Standard deviations in parentheses.*** Significant at greater than 1% level, ** Significant at 1% level, * Significant at 5% level.1 Each proposed agreement is estimated in isolation, but with existing agreements as in Section 3 in place. Coefficients on standard gravity

variables are omitted from the table.

18

significant) once openness variables are in-cluded (suggesting that the strong integrationreflects at least in part the entrepot role of Sin-gapore). The other two arrangements, betweenCanada and Japan, and Republic of Korea andMexico (both of which may be thought of asdefensive maneuvers given NAFTA and theprospects of an FTAA), show mixed results.In neither case is there any evidence of strongintegration between these economies overall (inboth cases, the coefficients are negative, butnot significant), nor for manufactures or serv-ices. In the case of agriculture, however, thereis a statistically significant pattern. There is avery strong positive bias in agricultural tradebetween Japan and Canada, and a very strongnegative bias in agricultural trade between Re-public of Korea and Mexico.

Given the uniformly negative estimatedcoefficients on the Republic of Korea-MexicoRTA dummy, there seems little doubt that suchan arrangement could not be described as natu-ral, irrespective of what definition is used. TheJapan-Canada estimates also provide little jus-tification for enthusiasm on natural trading blocgrounds, the positive coefficients on agricul-tural trade being a possible exception.

The estimated coefficients on the FTAA,which would bring together NAFTA and theeconomies of South and Central America, areinteresting. Here, merchandise is small withstatistically insignificant results. However,once the overall level of openness of theseeconomies is controlled (which is relativelylow), the level of integration appears stronglypositive. This might be interpreted as a case ofa “natural” trading bloc that, through lingeringprotectionism, has not yet become strongly in-tegrated in trade. As such, it perhaps providesa useful case upon which to judge the useful-ness of the natural trading bloc hypothesis.

The proposed Japan-Republic of Koreabilateral arrangement has negative coefficientsin all experiments except agriculture (positivebut not significant), indicating that these econo-mies are not strongly integrated. The coeffi-

cients on merchandise trade and manufacturesare significantly weak once the openness dum-mies are included. Hence there is little evi-dence that Japan and Republic of Korea form anatural trading bloc, despite their geographicalproximity. There may of course be many ex-planations for this (most obviously the politi-cal and economic rivalry between the two na-tions, and the associated bias in trade policy,such as Republic of Korea’s only recently aban-doned Import Sources Diversification pro-gramme).

Expanding the Japan-Republic of Ko-rea arrangement to include China results in apositive estimated coefficient on trade integra-tion between these economies, strongly signifi-cant in both manufactures and overall merchan-dise. Given the lack of integration betweenJapan and Republic of Korea, this might sug-gest that both are natural partners of China, andtherefore (if the natural trading bloc hypoth-esis is accepted) that a bloc centred on Chinamight be beneficial. Further expanding the ar-rangement to include ASEAN results in highlysignificant positive coefficients (smaller butstill positive and significant for agriculture).The addition of CER to the group results in aslight fall in the coefficients on manufacturesand overall, but raises those on agriculture.Thus the results do seem to lend support to thehypothesis of a “natural” trading bloc withinEast-Asia, although they can reveal little aboutthe potential welfare effects of such an arrange-ment. This is where CGE simulation can beuseful.

B. CGE simulation results and sensitivity

Using the GTAP model and the GTAP5data as described above, the implementationof the each proposed RTA was simulated in iso-lation by the complete removal of all tariffs ona preferential basis. The exception is APEC,which was provided with a benchmark. Here,the assumption is of MFN reform. Thesimulations are all comparative static, and thusemphasize efficiency effects in much the same

19

way as standard models of trade. The estimatedwelfare effects of each proposal are presentedin table 6 and 7, measured as the equivalentvariation in regional income, in $1997 millions(this is the estimated monetary equivalent ofthe change in consumption, evaluated at con-stant initial prices). Also presented are the es-timated approximate standard deviationsaround the welfare results, obtained using thesystematic sensitivity techniques developed byArndt (1996) and Arndt and Pearson (1998).

The results of CGE simulations areknown to be particularly sensitive to the as-sumed values of the Armington elasticities, andso it is to these values that the computed stand-ard deviations relate. We assume that the lowerlevel elasticities in the GTAP5 database are themean values of these parameters, and that thereare symmetric triangular distributions aroundeach of these parameters with minima andmaxima at mean ± 75 per cent. Each sectoralelement of the vector is assumed to vary inde-pendently. The assumption is maintained, thatthe upper level Armington parameters are dou-ble the lower level parameters. Because theArmington parameters enter the model as ran-dom variables, the model results are also ran-dom variables, and it is possible to use numeri-cal integration techniques to obtain approxima-tions of the means and standard deviations. Thestandard deviations, while only approxima-tions, allow to observe directly which resultsare robust to changes in the parameter values,and which are not. As a rule of thumb, if themean estimate maintains the same sign withintwo standard deviations, the probability is(more than 95 per cent) that the sign of the es-timated value is correct.

Considering the bilaterals first, the re-sults for which are presented in table 6, despitestrong evidence of trade integration betweenSingapore and Japan, the simulation resultsindicate negligible gains for either economy(actually small losses for Japan). The estimatedeffects, while small, do appear to be robust toparameter changes. By the same token, how-ever, the non-members are barely affected by

this agreement in net welfare terms, and thepotential for negative welfare effects on non-members as a consequence of trade diversionseems minimal. A similar result holds in thecase of Singapore–United States. The CGEtechniques do not indicate the presence of anysignificant net welfare gains (in fact the sum tomembers is negative). However, the (robust)welfare estimates vary widely over the twopartners, with Singapore estimated to suffer awelfare decline, and the United States a wel-fare improvement. Once again, there seemslittle evidence of substantial trade diversioneffects on non-members (this is the one prefer-ential simulation where the net-welfare of non-members is actually estimated to rise slightly).

The CGE simulations of a Japan-Canada agreement indicate a significant posi-tive welfare effect on both Japan and Canada,although it does not appear to be very robust toparameter changes in either case. Negativewelfare effects seem to be concentrated on theUnited States (and appear reasonably robust).Simulation of the other NAFTA-related bilat-eral, Republic of Korea–Mexico, reveals a simi-lar situation. There are insignificant and notvery robust positive welfare consequences ofsuch an arrangement for the members. How-ever, there is a negative welfare effect for theUnited States, which, though very small, doesseem to be robust. Hence, there may be somecause for concern within the United States re-garding the new bilateral agreements that itsNAFTA partners are negotiating within Asia.Decomposition of the welfare change (notshown) in both cases shows that the losses arealmost exclusively a consequence of terms oftrade declines, indicating that the effect is dueto a loss of preferential access to NAFTA part-ners. The estimated effects on other non-mem-bers are small (but generally negative).

The FTAA simulation indicates positivewelfare gains for all of the proposed FTAAmembers in the dataset. However, only in thecase of the United States is the estimated ef-fect clearly robust to parameter changes. Nega-tive net welfare effects for non-members indi-

20

Table 6. Estimated Welfare Effect of Proposed RTAs – Equivalent Variation($1997 millions)

SJ SUS JC KM FTAA

Mean SD Mean SD Mean SD Mean SD Mean SD

Argentina -0.1 (0.0) 1.1 (0.1) 10.1 (5.6) -0.5 (0.2) 975.0 (140.8)Australia -1.1 (0.2) 9.5 (1.8) -65.8 (15.3) -0.7 (0.3) -47.7 (2.6)Austria -0.5 (0.3) 14.7 (2.5) 2.2 (1.8) 0.3 (0.1) -21.9 (9.9)

Belgium 0.1 (0.3) 12.2 (2.0) -45.4 (17.7) 0.9 (0.3) -95.0 (25.9)Brazil -0.9 (0.2) 0.5 (3.5) -64.2 (14.0) -2.4 (0.8) 1 239.8 (570.7)Canada -0.7 (0.5) -22.0 (12.5) 4 604.5 (1 862.9) 4.9 (0.6) 95.4 (57.5)Chile -0.3 (0.1) -1.7 (1.1) -7.7 (1.8) -0.5 (0.1) 112.8 (19.2)China -3.8 (0.4) 25.4 (3.0) -63.9 (28.9) -25.9 (5.7) -228.6 (12.1)Columbia -0.1 (0.0) 1.7 (0.4) -5.9 (4.9) -0.4 (0.3) 325.3 (58.6)

Denmark -0.1 (0.2) 4.6 (1.2) 0.0 (1.4) 0.4 (0.0) -9.2 (6.3)Finland -0.2 (0.1) 3.5 (0.7) -0.9 (0.7) -0.2 (0.1) -23.2 (2.2)France -1.5 (0.6) 44.5 (6.8) 29.7 (10.3) 0.5 (0.2) -180.4 (30.0)Germany -3.8 (1.1) 62.8 (11.0) 55.8 (24.6) -9.2 (1.4) -491.6 (34.3)Greece 0.0 (0.0) 2.0 (0.6) 0.1 (1.4) -0.3 (0.0) -1.7 (1.4)Hong Kong, China -3.2 (0.8) 94.2 (18.0) 34.9 (17.4) -9.0 (1.0) -12.5 (1.9)

India -1.4 (0.3) 10.1 (1.2) 16.7 (6.0) -1.6 (0.6) -40.5 (2.1)Indonesia -1.4 (0.2) 11.1 (3.1) -0.6 (10.6) -4.1 (0.8) -45.5 (3.9)Ireland -0.9 (0.3) 17.4 (3.1) 2.1 (1.9) -0.4 (0.1) -29.3 (2.6)Italy -1.6 (0.8) 32.2 (5.1) -43.0 (19.5) -2.9 (0.6) -298.1 (38.9)Japan -33.8 (5.9) 23.2 (3.8) 306.0 (539.3) -2.8 (1.4) -610.2 (37.6)Republic of Korea 0.4 (0.4) 14.9 (2.7) 31.5 (17.2) 263.2 (62.5) -249.3 (28.8)

Malaysia 0.9 (1.1) 92.1 (11.5) 4.2 (6.8) -3.7 (0.3) -38.1 (1.7)Mexico -0.1 (0.1) -2.6 (2.5) 63.1 (31.0) 37.2 (19.1) 158.6 (109.7)Netherlands -0.3 (0.6) 25.0 (4.2) 3.4 (5.7) 1.1 (0.2) -102.8 (29.5)New Zealand -0.3 (0.1) 2.5 (0.5) -33.0 (11.8) -0.9 (0.3) -21.0 (1.4)Peru -0.1 (0.0) 0.0 (0.3) -12.6 (2.9) 0.0 (0.0) 129.6 (43.4)Philippines 0.5 (0.2) 17.5 (5.1) 7.0 (8.1) -2.8 (0.5) -8.2 (1.5)Poland 0.1 (0.1) 2.7 (0.4) -2.0 (1.5) -1.8 (0.2) 3.6 (3.6)

Portugal -0.1 (0.0) 3.0 (0.4) 2.8 (1.1) -0.1 (0.0) -11.9 (1.7)Rest of world -8.4 (2.2) 90.0 (16.8) -459.6 (144.6) -18.9 (2.8) -897.7 (78.9)Singapore 60.1 (16.9) -1 270.2 (198.3) 83.2 (38.5) -5.0 (0.9) -9.6 (12.6)Spain -0.5 (0.2) 14.0 (2.1) -5.3 (5.3) -2.5 (0.6) -201.3 (26.5)Sweden -0.2 (0.2) 7.3 (1.1) -1.1 (1.2) -0.2 (0.1) -70.8 (10.7)Switzerland -0.3 (0.0) 2.7 (1.0) -1.8 (2.0) 0.2 (0.2) -92.0 (12.6)

Thailand 1.1 (0.3) 31.3 (4.0) 50.3 (25.9) -2.2 (0.5) -24.5 (1.9)Turkey 0.1 (0.0) 2.0 (0.5) 7.6 (4.9) -0.3 (0.1) -15.0 (3.3)United Kingdom -1.2 (0.4) 31.5 (4.8) 14.1 (14.2) -1.5 (0.4) -224.7 (32.4)United States -3.3 (1.4) 1 009.3 (173.6) -1 840.2 (591.5) -151.1 (31.0) 3 823.9 (332.5)Venezuela -0.1 (0.0) 0.7 (0.3) -6.7 (2.4) -1.3 (0.3) 87.0 (62.9)

Members 26.2 -261.0 4 910.6 300.4 6 947.3Non-members -33.5 681.5 -2 240.9 -245.1 -4 098.6World -7.3 420.5 2 669.6 55.3 2 848.7

Source: Model simulations.

21

Table 7. Estimated Welfare Effect of Proposed RTAs – Equivalent Variation($1997 millions)

JK JKC A3 A3C APEC

Mean SD Mean SD Mean SD Mean SD Mean SD

Argentina 2.3 (0.3) -55.1 (3.4) -74.6 (5.4) -72.9 (3.2) 1215.1 (80.2)

Australia -57.2 (7.1) -260.8 (27.8) -493.0 (48.5) 3 940.4 (854.1) 2 797.0 (573.0)Austria 0.4 (1.6) 12.1 (12.7) 5.0 (23.7) 7.7 (27.8) -276.3 (125.4)Belgium -1.6 (2.3) -0.3 (19.6) -54.4 (39.7) -115.4 (60.8) -135.2 (38.6)Brazil -15.6 (3.5) -155.9 (27.1) -227.4 (38.2) -271.3 (38.6) 2 351.6 (442.3)Canada 1.0 (0.9) -50.8 (14.9) -21.0 (16.2) -159.1 (31.9) 2 032.9 (1 184.9)Chile -13.2 (1.8) -65.2 (3.5) -91.0 (5.3) -114.5 (7.8) 108.2 (92.4)

China -172.3 (27.6) 249.1 (721.2) 441.0 (651.4) 118.9 (463.1) 1 726.5 (1 800.3)Columbia -4.8 (1.0) -27.2 (6.2) -27.8 (6.9) -38.4 (7.4) 747.6 (42.5)Denmark -0.4 (0.9) -1.2 (7.1) -14.5 (13.7) -26.0 (19.6) 606.8 (360.0)Finland -3.5 (0.6) -24.7 (3.0) -62.4 (8.0) -69.0 (9.3) 29.6 (13.5)France 5.4 (4.2) 8.2 (48.5) -86.1 (79.8) -157.1 (106.5) 1 018.8 (173.3)Germany -60.1 (8.4) -398.6 (76.6) -803.6 (125.2) -984.3 (169.7) 1 849.4 (235.2)

Greece 0.5 (0.5) 12.0 (4.2) 19.3 (5.7) 17.4 (11.2) -273.3 (84.2)Hong Kong, China -8.9 (4.1) 2 811.1 (504.7) 3 410.0 (238.5) 3 389.2 (225.2) 6 788.5 (584.9)India -2.5 (1.7) -34.1 (18.5) -126.5 (34.5) -227.2 (39.9) 919.7 (155.5)Indonesia -54.5 (6.9) -251.9 (43.5) 621.8 (35.0) 420.6 (33.2) 734.9 (110.5)Ireland -14.0 (2.0) -38.3 (9.8) -64.3 (18.0) -68.4 (20.4) 84.8 (53.0)Italy -13.4 (4.6) -96.4 (39.6) -200.9 (69.6) -347.9 (92.1) 1 023.0 (117.7)

Japan 1 430.6 (153.9) 5 285.1 (809.9) 8 208.5 (1 080.5) 7 900.8 (1 535.5) 8 819.4 (2 380.8)Republic of Korea 291.8 (102.6) 5 535.2 (1710.9) 5 700.9 (1 736.8) 5 559.8 (1 742.8) 5 261.9 (1 902.7)Malaysia -53.7 (7.1) -248.1 (37.9) 182.7 (156.1) 72.8 (156.2) 94.2 (222.0)Mexico 5.0 (1.5) 33.3 (16.7) 62.4 (20.3) 34.9 (28.3) -1 036.4 (377.2)Netherlands -8.0 (3.3) -90.7 (29.2) -188.6 (58.3) -177.1 (71.6) 447.3 (292.9)New Zealand -5.4 (0.9) -101.3 (16.2) -141.6 (14.8) 1 484.1 (463.7) 1 301.5 (449.2)Peru -0.4 (0.3) -33.4 (5.4) -43.2 (6.5) -53.5 (8.2) 55.6 (78.9)

Philippines -22.6 (3.7) -96.5 (16.2) 22.9 (86.6) -108.4 (74.6) 747.4 (267.1)Poland 3.8 (1.1) 34.3 (9.7) 35.0 (13.9) 44.0 (15.2) 481.2 (77.3)Portugal 1.1 (0.5) 17.6 (5.0) 22.0 (6.4) 19.5 (9.0) 276.5 (56.2)Rest of world -372.4 (49.4) -2 731.2 (411.3) -4 335.9 (564.7) -5 468.9 (719.6) 1 527.7 (619.2)Singapore -30.4 (10.3) -135.7 (74.7) 116.2 (167.0) 158.6 (163.6) -1 183.2 (205.3)Spain 3.9 (2.4) 19.1 (21.6) -3.5 (34.1) -26.5 (45.9) 756.1 (55.5)

Sweden -5.1 (1.0) -46.3 (10.8) -101.4 (18.4) -121.2 (22.0) -237.0 (57.5)Switzerland -22.9 (3.6) -76.5 (19.0) -157.4 (31.1) -221.3 (47.7) 10.4 (8.9)Thailand -49.0 (7.8) -269.2 (49.9) 1 641.3 (374.3) 1 553.4 (364.9) 1 988.3 (574.9)Turkey 5.3 (0.7) 37.1 (9.7) 48.7 (12.7) 28.8 (20.0) -46.1 (115.6)United Kingdom -26.2 (8.9) -40.9 (64.9) -233.5 (104.8) -581.9 (208.4) 2 363.8 (525.1)United States -381.1 (48.1) -2 487.6 (287.6) -4 131.7 (383.4) -4 758.9 (441.9) 271.6 (1 119.6)

Venezuela -0.3 (0.2) -0.6 (3.2) 0.1 (3.5) -6.0 (4.6) -20.3 (4.8)

Members 1 722.4 13 880.5 20 345.3 24 069.5 30 508.4Non-members -1 370.9 -7 644.8 -11 491.7 -13 494.1 14 721.3World 351.6 6 235.7 8 853.6 10 575.5 45 229.7

Source: Model simulations.

22

cate that there is also considerable trade diver-sion occurs in this simulation.

Turning to the results for East-Asianintegration, which are presented in table 7.Scollay and Gilbert (2001) discuss the Japan–Republic of Korea agreement as a possible basefor formation of an East-Asian bloc. A situa-tion is envisaged whereby this agreement ex-pands to incorporate China (including HongKong,China), the economies of ASEAN andfinally the economies of CER (although noclaims are made as to the likelihood of theseconfigurations actually occurring). The Japan-Republic of Korea simulation alone results inrelatively small welfare effects for the partici-pants (although larger than the other bilateralsconsidered above). The results are robust forJapan, less so for Republic of Korea. Signifi-cant welfare losses to are estimated to be im-posed on non-members.

Expansion to include China substan-tially increases the total welfare gains, as wellas the gains to the members of the precedinggroup. The estimates for China are, however,relatively small and apparently highly sensitiveto the parameter assumptions (the Hong Kong,China results are larger and robust). The ex-tent of welfare losses to non-members also rise,but fall as a proportion of member welfaregains. Bringing in ASEAN to the group againresults in greater estimated mean net welfaregains to all existing members, and further wel-fare gains to the new members (although withthe exceptions of Indonesia and Thailand, theestimated gains to ASEAN members are notvery robust). Total member gains rise with thisexpansion also, as do non-member losses(though remaining roughly constant as a pro-portion of member gains).

The addition of CER results in a slightlydifferent pattern. The estimated welfare of thenew members rises (by a significant amountand with robust sign), but the estimated wel-fare of all the existing members falls (albeitonly slightly). Total welfare to members againrises, and total welfare to non-members again

falls. Given that the economies of CER areefficient agricultural producers, this result is alittle perplexing. Examination of the decom-position reveals that the previous members dogain from bringing in CER in allocative effi-ciency terms as a result of expansion of agri-cultural imports, as we would expect. It is smalldeclines in the terms of trade as a result of los-ing their preferential access that account for theslight welfare declines (relative to the RTAwithout CER).

The final simulation, of APEC on anMFN basis, provides a benchmark for evaluat-ing the size of the costs and benefits of the newRTA proposals. The estimated total welfaregains to APEC from MFN reform are largerthan those of any other group considered, ap-proached only by the other comprehensive EastAsian bloc considered. However, they are quitesensitive for some economies (notably Canada,China, Malaysia and the United States). Thechoice of MFN reform under the banner of“open regionalism” ensures that trade diversionis not a possibility.

C. Contrasting the techniques

The next task is to consider whether ornot the results of gravity model simulations andCGE simulations tell a consistent story in thecase of the new Asia-Pacific RTA proposals.If there is a consistency, as well as being ofanalytical interest, this would present moregrounds for confidence in the model predic-tions.

The bilaterals are begun with onceagain, the results for which are presented intables 5 and 6. From the gravity model, thetwo Singapore agreements were shown to havea high degree of integration. The CGE simula-tion techniques, indicate what intuition mighthave led some to expect – small bilateral agree-ments like these within the APEC region arelikely to have only marginal effects on the eco-nomic welfare of members and non-memberalike, positive or negative. Whether the ar-

23

rangements are “natural” or not in the gravitysense appears to be of little consequence inthese cases, although neither has any substan-tial trade diversion effect on non-members (thenet effect on non-members of Singapore–United States being positive, but small).

There is a clear consistency in the caseof Republic of Korea–Mexico. Here, the grav-ity model indicated that this was in no sense anatural trading bloc. The CGE simulation in-dicates negligible welfare gains (not evensignable in the case of Mexico) from a prefer-ential agreement, and also the presence of tradediversion. Hence, the gravity model and simu-lation techniques do appear to both identify thisagreement as an unlikely candidate for a modelRTA.

In the case of Japan-Canada, again thegravity model results do seem to coincide withthe predictions of the simulation techniques,at least in the context of agricultural trade. Thegravity simulations indicated a strong bias inagricultural trade, but little bias in other mer-chandise trade. The CGE simulations indicatedpositive (and large by the standards of the otherbilaterals) net welfare gains to both Japan andCanada. Harberger (1971) has shown that thewelfare effect of any change within a generalequilibrium system can be expressed in termsof its component parts by measuring changesin movements of goods across existing distor-tions. GTAP implements this decompositionautomatically using a set of routines developedby Huff and Hertel (1996). Examination of theresults (not shown), indicates that Japan’s wel-fare gain comes almost exclusively as a conse-quence of increased agricultural imports fromCanada. Canada’s gain, by contrast, comes al-most exclusively from improved agriculturalterms-of-trade with Japan (e.g. through in-creased exports). Hence, there is a consistencyof results across the two techniques in this case.However, we note that the estimation of netwelfare gains to the members does not rule outsignificant welfare losses to non-members inthe simulation approach, and so while there isa degree of consistency, the simulation tech-

niques do not support the supposition that natu-ral blocs are less trade-diverting.

For the FTAA, the gravity model didseem to indicate an intra-regional trade biasonce the overall openness of the economies wasincluded in the model. The natural bloc iden-tified in the gravity simulations did seem tocoincide with estimated positive net welfaregains for FTAA members. However, for manyof the economies the welfare gains are not veryrobust, and once again the results were not con-sistent with small welfare losses for non-mem-bers according to the simulation approach.

The East-Asian groups are considerednext (tables 5 and 7). It will be recalled fromthe discussion above that there was little evi-dence from the gravity approach of strong in-tegration between Japan and Republic of Ko-rea, stronger evidence of integration betweenthese two economies and China, stronger evi-dence still of integration between these threeeconomies and ASEAN and finally strong evi-dence of integration between all of the East-Asian economies and CER in terms of agricul-tural trade (though not overall).

The results of the CGE simulations areagain broadly consistent with the gravity re-sults. The Japan–Republic of Korea simula-tion alone, results in relatively small welfareeffects for the participants (at least relative tothe size of these economies), and welfare lossesto non-members. This fits with the gravity pre-diction that this is not a natural trading bloc.The addition of China/Hong Kong, China tothe group, results in a very substantial increasein the net welfare of the members, composedof increases for both the original members andthe new members. Again, this pattern matcheswith the increase in the degree of intra-regionaltrading bias (relative to Japan–Republic ofKorea alone) that the gravity approach captures.The pattern repeated with the addition ofASEAN, an increase in welfare for the newmembers and the old is observed under the CGEsimulation method, corresponding to the in-crease in the estimated RTA coefficient ob-

24

served under the gravity approach.

When the same grouping is simulatedwith the addition of CER, in the CGEsimulations there is again observed that totalwelfare gains to members increases, but in con-trast to the other expansions, the estimatedwelfare gain to existing members falls. Thegravity model simulations indicated a smallreduction in the overall degree of intra-regionalbias in this case on overall trade. On agricul-tural trade, there was an increase in the intra-regional bias, matching the welfare gains as-sociated with increased agricultural trade in thissimulation.

Thus the estimated welfare effects ofthese RTAs on the members of the RTAs domatch quite closely the results of the gravityanalysis – changing in line with the gravity pre-dictions. Do they lend support to the naturaltrading bloc hypothesis? The answer to thatquestion is less clear cut. To the extent that thenatural trading bloc hypothesis states that RTAsbetween “natural” partners are less trade divert-ing, the simulation results do not bear this out.In each case, the estimated trade diversion costs(in terms of welfare losses imposed on non-members) increases as the bloc becomes more“natural” according to the gravity definition.Moreover, since the pool of non-membereconomies is shrinking at each step, greaterlosses are being imposed on a smaller group of

non-members. It is difficult therefore to con-clude that the simulation results support thedesirability of “natural” blocs.

It may also be speculated that these re-sults support a hypothesis more basic than thenatural bloc. That is, they support the generalhypothesis that large blocs are better than smallones (at least for the members involved). Thereason is that the larger and more diverse thegroup of economies in the RTA, the more likelyit will be that one of them is an efficient pro-ducer of each commodity, and the less likely itwill therefore be that trade diversion will oc-cur. The results of the final simulation certainlysupport this hypothesis. The estimated totalwelfare gains to APEC from MFN reformdwarf those of any other group considered, formembers and non-members alike. Here thereform includes a large, diverse group of econo-mies. Moreover, the choice of MFN reformunder the banner of “open regionalism” en-sures that trade diversion is not a possibility.

Overall, the comparison of a gravitytype search for natural trading blocs, and thesimulation approach to estimating the effectsof proposed agreements seems to yield someconnection, but does not support the hypoth-esis that natural blocs are less likely to be tradediverting. If anything, the natural trading blocsmay be more likely to be welfare enhancingfor the members.

25

Both the gravity model and CGE ap-proaches suggest that there may be significantwelfare gains associated with some of the newRTA proposals in the Asia-Pacific region.These gains are largest when the group consid-ered is large and diverse. The results for mostof the bilateral agreements are somewhat lessimpressive. In many cases, there does appearto be a consistency between the “natural” blocestimates of the gravity approach and the simu-lation estimates of welfare benefits to mem-bers (but not clearly to an absence of trade-di-version).

Do the results imply that the new RTAsshould be actively promoted? Unfortunately,modeling work such as this cannot provide adefinitive answer to this question. It is clearthat the total benefits of liberalization within

APEC on an MFN basis are substantiallygreater than those associated with any of thenew RTA proposals. It is also clear from thesimulation work that even where the RTAs havesubstantial net benefits for the member econo-mies, they are likely to often impose substan-tial costs on non-members – including the re-maining members of APEC. The key questionis therefore the dynamic time-path question.Would the welfare costs imposed on non-mem-bers by these new agreements cause fractureswithin APEC that prevent the achievement ofthe Bogor goal? Or would they encourage con-glomeration of the disparate blocs withinAPEC, which combined with gradual elimina-tion of barriers to non-APEC members leadsto the same end-point as APEC MFN? Thiswill be an area of fruitful future research.

V. CONCLUDING COMMENTS

26

1 The authors consider the effect of the EuropeanUnion, NAFTA, AFTA, CER, MERCOSUR,the Andean Pact and APEC as part of their basescenario. They also consider the degree of in-tegration between Japan, Republic of Korea,China, AFTA, CER and other economies intheir analysis of new RTA proposals.

2 It should be noted that, because of the natureof the construction procedure of the GTAP da-tabase, the services data is not as “clean” as thetime series data used in the merchandise tradesimulations. Hence, the services results shouldbe interpreted cautiously.

NOTES

3 Modeling bilateral trade this way, when com-bined with transportation costs, is compatiblewith a gravity approach.

4 As this data is still a work in progress, thereremain some anomalies. Among the most glar-ing are high output and import taxes in themanufactures sectors in Singapore and HongKong, China. These were eliminated prior toour main simulations.

27

Armington, P.S. (1969). “A Theory of Demand for Products Distinguished by Place of Production”,IMF Staff Papers, 16:159–78.

Arndt, C. (1996). “An Introduction to Systematic Sensitivity Analysis via Gaussian Quadrature”,GTAP Technical 2.

Arndt, C. and K.R. Pearson (1998). “How to Carry Out Systematic Sensitivity Analysis via GaussianQuadrature and GEMPACK”, GTAP Technical 3.

Bhagwati, J. and A. Panagariya (eds) (1996). The Economics of Preferential Trading Agreements,AEI Press, Washington DC.

Bikker, J. (1987). “An International Trade Flow Model with Substitution: An Extension of theGravity Model”, Kyklos, 40:315–37.

Bora, B. and C. Findlay (ed.) (1996). Regional Integration and the Asia Pacific, Oxford UniversityPress, Melbourne.

Findlay, R. and A. Panagariya (1996). “A Political Economy Analysis of Free Trade Areas andCustoms Unions” in R. Feenstra, G. Grossman and D. Irwin (eds), The Political Economy ofTrade Policy: Papers in Honor of Jagdish Bhagwati, MIT Press, Cambridge.

Frankel, J.A. (1997). Regional Trading Blocs in the World Economic System, Institute for Interna-tional Economics, Washington DC.

Freund, C. (2000). “Different Paths to Free Trade: The Gains from Regionalism”, Quarterly Jour-nal of Economics, 65:1317–41.

Grossman, G. and E. Helpman (1995). “The Politics of Free Trade Agreements”, American Eco-nomic Review, 85:667–90.

Harberger, A.C. (1971). “Three Basic Postulates for Applied Welfare Economics: An InterpretiveEssay”, Journal of Economic Literature, 9:785–97.

Hengeveld, W.A.B. (1996). “World Distance Tables, 1948-1974”, ICPSR Database Reference 6152.

Hertel, T. (1997). Global Trade Analysis: Modeling and Applications, Cambridge University Press,New York.

Huff, K. and T. Hertel (1996). “Decomposing Welfare Changes in the GTAP Model”, GTAP Tech-nical 5.

REFERENCES

28

Kemp, M. and H. Wan (1976). “An Elementary Proposition Concerning the Formation of CustomsUnions”, Journal of International Economics, 6:95–97.

Krishna, P. (1998). “Regionalism and Multilateralism: A Political Economy Approach”, QuarterlyJournal of Economics, 63:227–51.

Krugman, P. (1991). “The Move Toward Free Trade Zones”, in Policy Implications of Trade andCurrency Zones, Proceedings of a Symposium sponsored by the Federal Reserve Bank ofKansas.

Panagariya, A. (2000). “Preferential Trade Liberalization: The Traditional Theory and New Devel-opments”, Journal of Economic Literature, 38:287–331.

Panagariya, A. and P. Krishna (1997). “On the Existence of Necessarily Welfare-Enhancing FreeTrade Areas”, Working Paper No.32, Center for International Economics, University of Mary-land.

Richardson, M. (1993). “Endogenous Protection and Trade Diversion”, Journal of InternationalEconomics, 34: 309–324.

Scollay, R. and J. Gilbert (2000). “Measuring the Gains from APEC Trade Liberalization: An Over-view of CGE Assessments”, World Economy, 23:175–197.

Scollay, R. and J. Gilbert (2001). New Regional Trading Arrangements in the Asia-Pacific?, Insti-tute for International Economics, Washington DC.

29

UNCTAD Study Series on

POLICY ISSUES IN INTERNATIONAL TRADEAND COMMODITIES

No. 1 Erich Supper, Is there effectively a level playing field for developing countryexports?, 2001.

No. 2 Arvind Pangariya, E-commerce, WTO and developing countries, 2000.

No. 3 Joseph Francois, Assessing the results of general equilibrium studies of multi-lateral trade negotiations, 2000.

No. 4 John Whalley, What can the developing countries infer from the UruguayRound models for future negotiations?, 2000.

No. 5 Susan Teltscher, Tariffs, taxes and electronic commerce: Revenue implicationsfor developing countries, 2000.

No. 6 Bijit Bora, Peter J. Lloyd, Mari Pangestu, Industrial policy and the WTO, 2000.

No. 7 Emilio J. Medina-Smith, Is the export-led growth hypothesis valid for develop-ing countries? A case study of Costa Rica, 2001.

No. 8 Christopher Findlay, Service sector reform and development strategies: Issuesand research priorities, 2001.

No. 9 Inge Nora Neufeld, Anti-dumping and countervailing procedures – Use orabuse? Implications for developing countries, 2001.

No. 10 Robert Scollay, Regional trade agreements and developing countries: The caseof the Pacific Islands’ proposed free trade agreement, 2001.

No. 11 Robert Scollay and John Gilbert, An integrated approach to agricultural tradeand development issues: Exploring the welfare and distribution issues, 2001.

No. 12 Marc Bacchetta and Bijit Bora, Post-Uruguay round market access barriers forindustrial products, 2001.

No. 13 Bijit Bora and Inge Nora Neufeld, Tariffs and the East Asian financial crisis,2001.

No. 14 Bijit Bora, Lucian Cernat, Alessandro Turrini, Duty and Quota-Free Access forLDCs: Further Evidence from CGE Modelling, 2001.

No. 15 Bijit Bora, John Gilbert, Robert Scollay, Assessing regional trading arrange-ments in the Asia-Pacific, 2001.