People s Republic of China s export threat to ASEAN: Competition in the US and Japanese markets

Transcription

1 ADB Insttute Dscusson Paper No. 2 People s Republc of Chna s export threat to ASEAN: Competton n the US and Japanese markets John Wess and Gao Shanwen Asan Development Bank Insttute January 2003 John Wess s Drector of Research at the ADBI. Gao Shanwen was then Research Assocate at the ADBI. The authors acknowledge the comments of Iwan Azs and Davd Roland Holst. The vews expressed n ths paper are those of the authors and do not necessarly reflect the vews or polces of the Asan Development Bank Insttute. 1

2 Introducton The ssue of People s Republc of Chna s (henceforth PRC compettve threat to ASEAN and other regonal exporters has receved consderable attenton n the aftermath of PRC s WTO accesson. Most modelng exercses avalable at present tend to suggest that welfare losses n the regon wll be relatvely slght snce all exporters wll gan from the expanson of world trade and the rse n mports nto the Chnese market (Ianchovna and Martn, However ths does not mean that there wll be no market dsrupton n the short-term to exporters n some sectors and some countres as a result of ncreased Chnese competton. Ths paper consders the ssue of competton by lookng at trends n mports nto the two of the man markets of the world economy Unted State and Japan. It addresses the queston - n what sectors was ASEAN losng compettveness to PRC pror to the WTO accesson and explores the economc characterstcs of these sectors. The perod of study s the second half of the 1990 s when competton from PRC ntensfed n a range of products. The analyss s conducted both for the man ASEAN exporters as a group and then moves to consder ndvdual economes. We focus on the larger ASEAN economes Sngapore, Malaysa, Thaland, Indonesa and the Phlppnes and exclude the smaller exporters such as Vet nam and Camboda. The mportance of the US and Japanese markets for these economes and PRC can be seen n table 1. The paper s n four man sectons. Secton 1 sets out a methodology for assessng the compettveness effect for ASEAN from PRC based on changng relatve market shares. It also gves the descrptve results from the applcaton of ths approach. Secton 2 sets out a regresson model, whch attempts to explan ths compettveness effect by characterstcs of dfferent export categores. Secton 3 gves the results of ths regresson analyss. We conclude wth some fnal observatons. Table 1 Share of US and Japan n total exports (-2000 for selected ASEAN economes and PRC. Share of US (% Share of Japan (% PRC Sngapore Malaysa Thaland Indonesa Phlppnes ASEAN (total for sx economes Source: unless otherwse stated all data are calculated from Unted Natons trade database (COMTRADE. 1. Measure of compettveness effect To establsh the degree of loss of market share to Chnese exports we apply a verson of constant market share analyss. 1 growth for a gven commodty to a partcular market (e.g. US or Japan can be decomposed nto a share effect (assumng country j keeps a constant share of the market and compettveness effect (allowng for ts changng market share. Hence we have 1 For a dscusson and applcaton of constant market share analyss see ADBI (2002. For a conceptual dscusson, see Rchardson (1971. The approach here uses a verson of equaton (6 from Rchardson (1971. The analyss dffers from a conventonal constant market share approach as t looks separately at two markets (the US and Japan and at dfferent export categores. Hence the commodty composton and export composton effects, whch are relevant when total exports to the world are the focus of attenton, are not requred here. 2

3 Xj = X 1 j + X 2 j (1 where X s exports and s the absolute change n. X 1 = Q.sj (2 where Q s total mports of n the market concerned (at the end of the perod and sj s the ntal market share of country j n mports of and X 2 j = sj.q ( sj/sj - sk/sk + sk/sk. sj.q (3 where now compettor country k has been ntroduced, so sk s k s market share for product. Equaton (3 reduces to the dentty X 2 j = Q. sj, that s compettveness equals ntal mports tmes the change n market share. However expressng the compettveness effect n (3 allows the ntroducton of a comparator compettor economy (n ths case PRC. The two dfferent terms n (3 gve dfferent nformaton. The frst term gves the absolute change n exports from j due to the change n ts market share vz-a-vz PRC (that s the due to the change n ts compettveness relatve to PRC. The second term gves the absolute change n exports from j due to the change n PRC s market share relatve to the rest of the world (ncludng j. The sum of these two effects s the total compettveness effect for j. Addng equatons (2 and (3 n (4 allows a decomposton of the change n exports of for country j, so that Xj = Q.sj + sj.q ( sj/sj - sk/sk + sk/sk. sj.q (4 A negatve sgn for the second term ndcates a loss of compettveness vz-a-vz PRC ( sj/sj - sk/sk<0; however ths may be compensated by the movement of PRC s compettve poston vz-a-vz the rest of the world, sk/sk. Countres for whch the sum of these two effects s negatve wll be losng compettveness. Whether overall exports fall wll then depend on the strength of the general demand effect, whch s the frst term ( Q.sj. Applcaton to ASEAN Here we apply ths approach to the exports of fve man ASEAN economes to the US and Japan. We take as our perod of analyss as dsaggregate data are avalable for ths perod from the UN trade database. We proceed n stages commencng ntally wth an analyss at the 2-dgt SITC level, before movng to a more dsaggregate analyss at the fourdgt level. Table 2 shows the decomposton, for ASEAN as a whole, for the fve 2-dgt SITC categores n whch the loss to PRC s greatest. Table 2 ASEAN compettveness decomposton US market: key SITC categores US$ thousand SITC 75 s ncrease to US - Constant Market share Overall compettveness Compettveness vz PRC ncrease as % of 3

4 00 effect exports % of export ncrease SITC 77 s ncrease to US - 00 Constant Market share effect Overall compettveness Compettveness vz PRC % of export ncrease SITC 76 s ncrease to US - 00 Constant Market share effect Overall compettveness Compettveness vz PRC ncrease as % of total exports ncrease as % of exports % of export ncrease SITC 89 s ncrease to US - 00 Constant Market share effect Overall compettveness Compettveness vz PRC % of export ncrease SITC 82 s ncrease to US - 00 Constant Market share effect Overall compettveness Compettveness vz PRC ncrease as % of exports ncrease as % of exports % of export ncrease In each case there has been a strong effect from the loss of market share relatve to PRC. For SITC 77 ths was compensated by a gan n market share relatve to other exporters. For the others the mpact of the overall growth of mport demand was large enough for there to be a bg rse n exports n SITC 75, 77 and 82 and a more modest ncrease for SITC 76 and 89. In none of these major SITC was there an absolute declne n sales to the US, although there was a relatvely rapd eroson of market share to PRC. Table 3 shows a smlar decomposton for the Japanese market. There gven the greater mportance of prmary and resource-based products, SITC 34 (Natural Gas, 03 (Raw Fsh and 63 (Cork and Wood are ncluded n the fve 2 dgt categores n whch loss relatve to 4

5 PRC s greatest. As for the US, SITC 75 and 76 are n ths group. In the Japanese market for these fve man SITC categores there s no gan relatve to other exporters, and n relaton to SITC 03 and 63 there was an absolute declne, wth 76 havng only a relatvely modest ncrease. Table 3 ASEAN compettveness decomposton Japanese market: key SITC categores-00 US$ thousand SITC 34 s ncrease to US - 00 Constant Market share effect Overall compettveness Compettveness vz PRC % of export ncrease SITC 75 s ncrease to US - 00 Constant Market share effect Overall compettveness Compettveness vz PRC % of export ncrease SITC 76 s ncrease to US - 00 Constant Market share effect Overall compettveness Compettveness vz PRC % of export ncrease SITC 03 s ncrease to US - 00 Constant Market share effect Overall compettveness Compettveness vz PRC % of export ncrease SITC 63 s ncrease to US - 00 Constant Market share effect Overall compettveness Compettveness vz PRC % of export ncrease ncrease as % of exports ncrease as % of total exports ncrease as % of exports ncrease as % of exports ncrease as % of exports 5

6 2. Regresson analyss and methodology Here we report the results of our dsaggregate analyss that ams to establsh n more detal the type of products n whch ASEAN compettveness vz-a-vz PRC s beng eroded n the US and Japanese markets. In terms analyzng the change n compettveness we use as our measure of compettveness, change n compettveness relatve to PRC, whch s the second term n equaton (4, scaled by dvson by total exports n n the same category. We calculate ths measure for both markets, separately. Usng the notaton n equaton (4 compettveness (COMP s measured as COMPj = [ sj.q ( sj/sj - sk/sk ] / Xj where Q s total mports of n the market concerned (at the end of the perod and sj s the ntal market share of country j n mports of ; sk s the market share of the compettor economy k (n ths case PRC, denotes change n and Xj s ntal exports of from j to the market concerned. Where there s a gan n market share relatve to PRC COMP wll be postve and where there s a loss t wll be negatve. Ths varable s related drectly to the more conventonal measure of compettveness - the change n the revealed comparatve advantage rato (RCA. When the latter s defned (as below as the relatve rato for a country n a partcular market wth PRC as a comparator, then rankng by our compettveness measure wll be dentcal to rankng by the change n RCA. In the analyss we test the hypotheses that - loss of compettveness s systematcally related to the characterstcs of trade categores, whether n terms of technologcal characterstcs, or patterns of specalzaton; - loss of compettveness dffers systematcally between countres n the ASEAN group; - loss of compettveness trends are smlar between the US and Japanese markets. By the logc of constant market share analyss we would expect common trends between dfferent markets, as exporters wll be exposed to common supply-sde shocks n producng for dfferent markets. To test these hypotheses we utlze a smple model that makes compettveness (defned as above as the export gan vz PRC as a proporton of ntal exports a functon of the characterstcs of products, general shfts n compettveness and changes n tastes as a demand factor. We use a measure of specalzaton (RCA at the start of a perod to explan changng compettveness over ths perod. We use the ntal RCA as one of the explanatory varables for change n compettveness, on the grounds that the ntal RCA can be taken as a proxy for the relatve output level and factor ntensty of dfferent products. 2 In prncple RCA can be related to the change n compettveness through two possble routes: 2 Because we work at the dsaggregate 4 dgt level the model s very constraned by data avalablty, and because of our hghly dsaggregate focus we do not have access to conventonal measures that can explan changes n export market shares, such as unt labor cost or R and D actvty. Hence the use of the ntal RCA varable to capture the ndvdual characterstcs of products. Analyses that utlze such varables must normally work at much hgher levels of aggregaton; for example Carln et al (2001 for OECD countres focus on changes n market share at the level of 12 dvsons of manufacturng. 6

7 a though shocks that affect relatve factor prces and ther mpact on dfferent commodtes dfferently n lne wth ther factor ntensty n ths route the sgn on the RCA term wll be ambguous varyng wth the nature of the prce shock; b through a catch-up effect, so that where a compettor has low ntal market shares t wll be easer for them to gan n actvtes n whch an economy s hghly specalzed n ths route the expectaton s that the RCA term wll always be negatvely related to change n compettveness. Ths wll be most relevant for very large and very small observatons for RCA snce, f catch-up s a unversal pattern across all observatons, ultmately t wll result n homogenety n export structure across countres, whch s mplausble. Our theoretcal model focuses on the frst of these two routes, although t must be acknowledged that emprcally, whenever a negatve relaton between the compettveness and RCA varables s found, we wll be unable to dstngush between whch of these two mechansms s at work. Formally for product and a gven country j we have Comp j = f ( MC j, D j where dcomp j / dmc j <0, MC j s margnal cost for n country j relatve to ts cost n compettor economy k, and D j s a demand factor for j s exports, such that a change n D j reflects a shft n taste relatve to exports of from k. In turn MC j = f (Q, FP, T, Z, where Q s the relatve output of and FP s the relatve factor prce rato (captal rental/wage for countres j and k. T refers to a unque technology parameter specfc to an ndvdual product and Z s a general relatve compettveness term for the two economes reflectng supply factors that are ndependent of factor ntenstes. FP s the rato (r k/ r l, where r k s the opportunty cost of a unt of captal and r l s the wage. 3 We make the assumpton that due dmnshng returns to fxed factors margnal costs rse wth output, so that dmc / dq > 0. We do not have product specfc data on Q hence we apply as a proxy measure, relatve revealed comparatve advantage (RCA, whch can be calculated readly from the trade database Relatve revealed comparatve advantage s defned as RCA = (X j /X tj /(X k /X tk where X refers to export value, t stands for total exports and k s the comparator economy. Snce a rse n RCA ndcates that the commodty concerned s a growng share of trade (and output, by assumpton MC wll rse wth RCA. RCA wll be drectly related to the relatve factor ntensty of ndvdual products. Hence products ntensve n an economy s relatvely abundant factor wll have a hgh RCA; e f ASEAN s captal- abundant relatve to PRC, captal-ntensve products can be assumed to have a hgh RCA n ASEAN. A formal proof of the dervaton of our basc estmaton equaton s gven n Appendx 1. We show n Appendx 3 In prncple the model can be based on any two factors of producton. We choose the conventonal measures of captal and labor, but other combnatons such as sklled and unsklled labor or sklled labor and land could also be used. The crtcal assumpton s that the relatve prces of the two factors selected determne specalzaton on the bass of the factor-ntensty of products; see Wood (1997 for a dscusson of the underlyng theory. 7

8 1 that n our smple model the sgn on the RCA term s determned by the relatve change n the captal-rental to wage rato n an economy and ts comparator and captal and labor endowments n the two economes at the start of the perod of study. For an economy n whch there s both a fall n the captal rental to wage rato relatve to a comparator and a lower ntal endowment of captal relatve to labor the sgn on the RCA term must always be negatve. In ths smple model n the long-run n an equlbrum stuaton there can be no systematc relatonshp between RCA and compettveness, snce RCA cannot grow or shrnk ndefntely. At any pont n tme an economy can be at a transton from one equlbrum to another, hence a relatonshp may be found. The emprcal analyss focuses on ths transtonal stuaton. In the model by assumpton there wll be a postve relaton between the relatve margnal cost lne for countres j and k (MC and RCA (for a gven technology parameter and gven tastes. However the MC lne can shft due to changes n compettveness as a result of ether product specfc or country factors. The pattern of these changes wll determne the relatonshp between the ntal RCA and changes n compettveness over a gven perod. For an ndvdual commodty fgure 1 llustrates the logc nvolved. As we assume dfferentaton of products between supplers, for product there wll be dfferent prces between exportng countres j and k. The horzontal axs of fgure 1 shows both the relatve prce rato (Pj/Pk and the relatve margnal cost rato (MCj/MCk. Shfts n MC reflect a gan n compettveness for one country. In fgure 1 startng from pont A a shft from MC to MC creates a compettveness gan for country j (snce at pont A MCj/MCk<Pj/Pk exports from j wll rse relatve to compettor k and ths allows country j to capture market share at the expense of ts compettor over the transton from A to B. Startng from pont B an unfavorable shft for j s the move from MC to MC and ths leads to a compettveness loss and thus a transton to A. The new equlbrum s where there s equalty between the relatve world market prce rato (P and the relatve MC lne. As our analyss focuses on the transton we do not ncorporate feedback to the ntal RCA from the compettveness change over the perod. For a favorable shock for j, there wll be fall n MC. If ths fall s proportonately greater the hgher s the RCA ths wll lead to a postve relatonshp between RCA and compettveness. Ths the case llustrated n the shft from MC to MC n fgure 1. If the change n MC s proportonately greater at a lower level of RCA then the reverse holds and there wll be negatve relatonshp between change n compettveness and the RCA. An unfavorable shock mples an upward shft n MC. Where ths rse n cost s proportonately greater the hgher s the RCA (as n the move from MC to MC n fgure1 loss n compettveness wll be greater the hgher the RCA and the relatonshp wll be negatve. Smlarly f the rse n MC s proportonately lower at hgher levels of specalzaton losses wll be less the hgher s the RCA and the relatonshp wll be postve. Changes n MC can be related to changng relatve factor prces by the factor ntensty of dfferent products. A favorable factor prce shock for ASEAN mples a fall n the prce of ether captal or labor, relatve to that n PRC. Conversely an unfavorable prce shock mples a rse n one of these prces relatve to PRC. Where there s a favorable captal prce shock that lowers the captal rental to wage rato n an economy that s abundant n captal, ths wll lead to a fall n MC that s proportonately greater for products wth a hgh RCA, whch by assumpton wll be relatvely captal ntensve. A smlar effect wll occur for a favorable wage prce shock that rases the captal-rental to wage rato n economes that are labor-abundant. In the model because of the focus on the relatve poston these changes n factor prce ratos must be relatve to the comparator economy. Where MC lnes move n parallel they wll be unrelated to RCA and change wll be determned by a general shft n compettveness that s neutral as regards factor ntensty. 8

9 Relatve MC / Relatve Prce MC MC P A B RCA Fgure 1: RCA and Compettveness. The poston of ASEAN countres and PRC n ths model s exactly the reverse, as compettveness s n relaton to PRC and MC and RCA are relatve to PRC. Hence an unfavorable change for an ASEAN country s a favorable change for PRC. If ths change for an ASEAN country s greater the hgher the RCA, ths also means the reverse for PRC; that s the change s greater the lower the RCA for PRC. In general we expect ASEAN economes to be relatvely abundant n captal relatve to PRC. Ths proposton s supported by the fact that n, at the start of our perod, ncome per capta n all of the ASEAN economes covered here was hgher than n PRC both n calculatons n current prces at actual exchange rates and n purchasng power party estmates. Hence on the bass of ths assumpton n ASEAN, products wth a hgh relatve RCA are lkely to be relatvely captal ntensve, whlst n PRC those wth a hgh RCA are lkely to be more labor-ntensve. How shfts n relatve factor prces wll be related to RCA wll thus depend on whether the captalrental-wage rato rses or falls. For example, a favorable captal prce shock for PRC means a cheapenng of the relatve cost of captal for PRC and ths wll convert nto an unfavorable shock for ASEAN and a rse n ts MC, whch s greater n ts more specalzed products (as the move from MC to MC n fgure 1. The mpled negatve relatonshp between compettveness change and RCA for ASEAN s what s observed n our subsequent analyss. 4 In terms of the expresson for FP above a favorable prce shock for PRC can be nterpreted as a fall n the opportunty cost of captal n PRC relatve to that n ASEAN, for example due to relatvely hgher domestc savngs or FDI nflows. As a test of the basc model we also check the results for Inda, an economy wth a lower ncome per capta than PRC n by any measure, and by mplcaton an economy wth a lower captal-labor endowment. If some extreme outler observatons for the RCA term are excluded, on the grounds that catch up effects are lkely to be stronger for these (we nclude the RCA terms from 0.1 to 10, we fnd that for the US market the sgn on the RCA term for Inda s sgnfcant and the opposte of that for ASEAN as a group and all ndvdual ASEAN countres; t s postve for Inda as compared wth negatve for ASEAN. Ths s what s 4 A cheapenng of captal n PRC relatve to compettor economes wll be expected to ncrease captal ntensty, whch wll n turn rase labor productvty. For the perod (more recent years are not avalable at the three dgt level of manufacturng, real labor productvty growth n PRC exceeded that n all ASEAN countres covered here n 90% of the cases, whch s consstent wth ths hypothess. Productvty fgures are calculated from the World Bank Trade and Producton database (see Ncta and Olarreaga,

10 predcted by our smple model, f mproved compettveness s due to a fall n the relatve cost of captal vz-a vz labor n PRC. Ths wll mean that n relaton to Inda the latter s loss of market share wll be lower n ts more specalzed, that s ts labor-ntensve, products (for a further dscusson see Appendx 1. Compettveness behavor may also be nfluenced by the speed of growth of the market, so we also try as an ndependent varable the growth of the US market for the category concerned. It should be easer to mantan a market share aganst a new entrant n a fast growng market, as both a country and ths new entrant can grow at the expense of thrd party compettors, so a postve sgn s expected. However ths varable s never sgnfcant and s not reported. In addton we also experment wth other possble varables. We calculated unt value ndces for PRC and ASEAN and attempted to use dfferences between these as proxes for changes n relatve prces. However unt values are only avalable for a lmted number of trade categores and hence we are unable to apply ths prce varable consstently. We also tested for the mpact of the rato of exports from ASEAN n partcular trade categores to exports from PRC n that category at the begnnng of the perod. However ths varable s closely correlated wth the RCA measure and was dropped due to mult-collnearty. Mathematcally the change n compettveness wll be related to RCA by a concave lne; CC n fgure 2 llustrates the case where the coeffcent on RCA s negatve. Ths means that there wll be a crossover pont for the RCA at whch a product moves from compettve to uncompettve. We only report these ponts where the ndependent varables n the model are all sgnfcant and the RCA value at the crossover pont s above a mnmum fgure of There are a small number of cases where these condtons are met. Although t complcates the model we are forced to ntroduce changng demand condtons nto the analyss. Conventonal constant market share analyses omt any demand term and assume that supply changes that affect compettveness determne changes n a country s share n a gven market. However, our emprcal observatons of behavor n the US and Japanese markets fnd that for a sgnfcant number of trade categores the sgn on the compettveness term s the opposte; n other words a good may be ganng n relaton to PRC n one market and losng n the other. Assumng, as we do, that there s no sgnfcant change n relatve transport costs or n relatve trade barrers facng exporters, ths mples that products are dfferentated and non-homogeneous and that, provded the changes are not random, they must be caused by demand shfts. We try to capture ths effect by ncludng a dummy for products subject to demand shfts due to dfferental taste change n the two markets. In fgure 1 ths mples that for such products the horzontal prce lne wll move up (for a favorable demand shft or down (for an unfavorable one. 5 5 The general constant term (a1dvs explaned below pcks up both neutral shfts n compettveness unrelated to factor ntensty and any addtonal taste changes that occur that are n the same drecton n the two markets. 10

11 COMP C RCA C Fg 2. Illustraton of RCA and compettveness change As total trade covers a wde varety of product types to mpose some pattern on the data we classfy these by the nne product categores of all ( These can be nterpreted as descrptve classfcatons of relatvely homogeneous product categores. Our use of dummy varables reflectng these nne categores mples that there s broad homogenety wthn each n terms of the response of dfferent products to the explanatory varables. However, wthn these classfcatons we also attempt to capture the degree of technologcal sophstcaton by applyng the sophstcaton ndex of wan (2002. The wan ndex s calculated at the 4-dgt level to capture the technologcal sophstcaton of a trade category. It s based on the assumpton that the sophstcaton of a commodty can be approxmated by the weghted average ncome per capta of ts exporters wth the weghts determned by the share of each exporter n the US market. Ths ndex gves a unque value for each trade category. The expectaton s that products based on assembly type producton wll have a relatvely low value of the ndex (as they are assembled n low wage economes and that products wth hgher local value-added wll have a hgher value of the ndex. If t s the case that ASEAN s market poston s protected by a technologcal advantage wthn a broad trade category, we expect that the sgn on the wan ndex would be postve. 7 Summary of the regresson model To summarze the regresson analyss takes the form Compt = ( DVSt + ( β β 9 DVDt + ϕadt + ( γ γ 9 + ηadt og( RCA + ( δ δ 9 TECH + ε t where COMP s the compettveness term relatve to PRC as a proporton of exports, TECH s a measure of technologcal characterstcs of ndvdual products (at the SITC four dgt level, RCA s the relatve revealed comparatve advantage measure for the same 6 The all classfcaton s a wdely cted approach, whch utlzes earler classfcatons and data on R and D expendture, to classfy trade categores by factor and technology characterstcs. ke any such broad approach t s subject to reservatons concernng the homogenety of the categores. It groups products at the 3 dgt SITC level nto the followng nne man classfcatons prmary products, resource-based manufactures, low technology manufactures (splt between textles, garments and footwear, and others, medum technology manufactures (splt between engneerng, automotve products and process products and hgh technology (splt between electroncs and electrcals and others. 7 Ths expected sgn s also confrmed by the predcton of our smple model as dscussed n Appendx A.1, although there the role of the wan ndex s ratonalzed n a slghtly dfferent way as a proxy for relatve productvty levels between compettors. 11

12 product at the 4-dgt SITC level, whch s defned as a country s conventonal revealed comparatve advantage dvded by that of PRC for the same product. Snce we expect that the change n compettveness s a structural adjustment process n response to some external shocks as explored n the appendx and that the adjustment wll not be completed n the short sample perod, we use the RCA calculated at the startng pont of the perod to capture the structural characterstcs of products, whch we assume wll not change fundamentally over our sample perod. The same can be sad for the wan technology ndex that s a contnuous varable used for TECH. In addton to allow for other smlar characterstcs of the trade categores, products are grouped nto the nne categores used n the all classfcaton to reflect non-technologcal smlartes, whch n the above equaton s represented by the nne coeffcents for the same explanatory varable. These are reflected n nne product category dummes. The assumpton s that wthn each category all trade observatons wll behave n the same way n response to both demand sde shocks (that s taste change and supply sde shocks (general compettveness shfts unrelated to specalzaton and technologcal sophstcaton. However, to dstngush products where there have been demand sde shocks, defned as those where the sgn on the compettveness change dffers between the US and Japanese markets, we ntroduce two further dummes. A demand shock dummy (DVD takes a value of unty when the sgn on the compettveness term between the two markets dffers and zero when the sgn on the compettveness term s the same. The supply shock dummy (DVS takes a value of unty when the sgn n the two markets s the same and zero when t s opposte. The constant DVS refers to the nteracton term between a dummy for each of these nne groupngs and DVS. It can be nterpreted as pckng up the mpact of general supply shocks, unrelated to ether factor ntensty or technologcal sophstcaton, plus generally favorable demand factors that are unform between the two markets. 8 The constant βdvd refers to the nteracton term between a dummy for each of these nne groupngs and DVD. It should pck up the mpact of changes n taste that vary between the two markets. In addton the terms γtech and δrca are nteracton terms between the TECH and RCA varables and the dummes for the nne product groupngs. We use a log form for δrca on the grounds that any relaton between compettveness change and RCA s unlkely to be lnear and ths gves us the concave relaton n fgure 2. In ths way we attempt to test for the mpact of specalzaton and technologcal sophstcaton on compettveness, whlst controllng for changes n tastes as between the two markets. In addton, to allow for the possble dsruptons brought about by the Asan fnancal crss and ts aftermath, we ntroduce further annual dummes (AD that take a value of 1 for , and , separately. We specfy the equaton so that the Crss may affect the constant term as well as the coeffcents for RCA smultaneously, but not the technologcal varables. 3. Regresson Results We conduct the ntal analyss across 690 four-dgt SITC categores frst for the ASEAN countres as a group and then for each ndvdually. We use a weghted least squares approach, because from the nature of the dependent varable we know there wll be heterogenety n the error terms. In examnng the perod we apply three alternatve approaches; we take the averages of all varables over the perod, we take three year movng averages and (as explaned above we take annual observatons wth dummes for the years , and to pck up the effect of the Fnancal Crss and ts aftermath. The results are consstent between these three approaches, however as 8 Any changes n relatve trade barrers common to both markets would be pcked up here, whlst changes that vary between the two markets would be pcked up by DVD. 12

13 the latter usng annual data gves both the maxmum number of observatons and the largest number of sgnfcant coeffcents, we report ths verson of the analyss. US Market Table 4. US Market: Regresson results for ASEAN and ndvdual countres (annual data Country ASEAN Indonesa Phlppnes Malaysa Thaland Sngapore I (0.4 (0.6 (1.2 (1.0 (0.5 (1.0 II (0.3 (0.5 (0.8 (0.9 (0.4 (1.1 III (0.2 (0.2 (0.4 (0.6 (0.3 (0.5 IV (1.6 (5.4 (14.2 (7.8 (3.1 (3.6 Constant V (0.2 (0.3 (0.5 (0.3 (0.3 (0.3 VI (0.3 (0.6 (0.5 (0.4 (0.3 (0.5 VII (0.2 (0.3 (0.4 (0.4 (0.2 (0.9 VIII (0.8 (1.2 (4.5 (1.3 (0.9 (4.1 IX (0.5 (4.2 (1.4 (1.1 (0.9 ( (0.02 (0.03 (0.03 (0.02 ( (0.02 (0.03 (0.03 (0.02 ( (0.02 (0.03 (0.02 (0.02 (0.03 RCA I term (0.03 (0.04 (0.05 (0.04 (0.02 (0.03 II (0.02 (0.03 (0.03 (0.02 (0.02 III (0.02 (0.02 (0.02 (0.03 IV (0.09 (0.1 (0.2 (0.3 (0.08 (0.2 V (0.02 (0.02 (0.02 (0.02 VI (0.009 VII (0.02 (0.03 (0.02 (0.03 VIII (0.03 (0.03 (0.08 (0.04 (0.04 (0.05 IX (0.03 (0.2 (0.06 (0.07 (0.06 (

14 Annual dummy Annual dummy Annual dummy ( ( ( ( ( ( ( I (0.4 (0.6 (1.2 (1.0 (0.5 (1.0 II (0.3 (0.5 (0.8 (0.9 (0.4 (1.1 III (0.2 (0.2 (0.4 (0.6 (0.3 (0.5 IV (1.6 (5.3 (14.2 (7.8 (3.1 (3.6 Demand V dummy (0.2 (0.3 (0.5 (0.3 (0.3 (0.3 VI (0.3 (0.6 (0.5 (0.4 (0.3 (0.5 VII (0.2 (0.3 (0.4 (0.4 (0.2 (1.0 VIII (0.8 (1.1 (4.5 (1.3 (1.0 (4.1 IX (0.5 (4.2 (1.5 (1.1 (0.9 (1.6 I (0.05 (0.07 (0.1 (0.1 (0.06 (0.1 II (0.03 (0.05 (0.09 (0.1 (0.04 (0.1 III (0.02 (0.03 (0.05 (0.08 (0.03 (0.06 IV (0.2 (0.6 (1.4 (0.8 (0.3 (0.4 wan V Index (0.02 (0.04 (0.05 (0.03 (0.04 (0.03 VI (0.03 (0.06 (0.06 (0.04 (0.04 (0.05 VII (0.02 (0.03 (0.05 (0.05 (0.02 (0.1 VIII (0.08 (0.1 (0.5 (0.1 (0.1 (0.4 IX (0.05 (0.5 (0.2 (0.1 (0.1 (0.2 Observatons GS/OS WS WS WS WS WS WS Adjusted R- squared 27.2/ / / / / /-0.9 Notes: 1., and denote sgnfcance level of 1%, 5% and 10% respectvely. 14

15 2. In the last row the frst number s the Adjusted R-square for weghted statstcs and the second one s for un-weghted statstcs. Coeffcents and standard errors (n parenthess are rounded to the frst non-zero number. 3. To exclude data exceptons, nstead of employng data dummes we ntroduce a flter condton for the dependent varable whlst the cutoff s set at 2 (or 5 n some cases. At ths level the degree of freedom s not reduced very much, but the heterogenety of observaton fluctuaton has been substantally reduced. An ncrease of ths cutoff to 20 does not alter our results above. 4. Notably for most equatons here the error terms do not follow a normal dstrbuton. 5. The regresson equaton takes the form of CG = C(1 + C(2 log( RCA C(3 kwan + C(4 demand dummy + C(5 data dummy We have also tred the lnear form of the RCA term where the results are broadly the consstent. The addton of here s to allow for the observatons where RCA s zero. Ths sem-log verson s slghtly superor n term of adjusted R-squared and number of sgnfcant cases. Sector classfcaton and number of observatons: (accordng to the classfcaton n all (2000, revsed to verson 3 of SITC Type Maxmum sze Sector I 170 Prmary product II 254 Resource-based manufactures III 122 Textle, garment and footwear IV 34 Automotve V 166 Engneerng VI 50 Electronc and electrcal VII 110 Other low tech VIII 87 Process IX 32 Other hgh tech In analyzng the results from table 4 the followng man ponts emerge. 1. For ASEAN as a whole there s evdence of a statstcally sgnfcant loss of compettveness n fve of the nne product categores. These are the categores of resource-based manufactures, engneerng products, electroncs and electrcals, the other low technology category and other hgh technology goods. For all of these the loss s sgnfcantly greater the hgher s the degree of specalzaton as measured by the RCA, so that for these good losses of market share relatve to PRC are found consstently n the more specalzed areas. Ths appears to be a general pattern snce for ASEAN as a whole the sgn on the RCA term s always negatve except for the case of textles and garments, where t s postve but nsgnfcant. Technologcal sophstcaton, as measured by the wan ndex, s sgnfcant wth a postve sgn for the categores engneerng and other low technology goods, mplyng that losses are less n the more technologcally advanced products. For textles and garments the sgn on the wan ndex s negatve and weakly sgnfcant mplyng the reverse, although ths category appears to be nfluenced by some favorable demand shfts. There s no systematc tendency for gan of compettveness for ASEAN as a whole n any of the product categores, although as we dscuss below there s some lmted evdence for ndvdual countres. It thus appears that although there s a strong tendency to loss of market share relatve to PRC, ths s not across the board and the general hypothess of dfferental compettve effects by product category s 15

16 supported. osses are found across a wde spectrum of actvtes; n both hgh technology (electroncs and electrcals and other hgh technology products, medum technology (engneerng goods, relatvely smple products (other low technology, as well as n resource-based manufactures. There s no evdence of systematc loss of market share for ASEAN as whole for prmary products, automotve products and process products and only very weak evdence for textles and garments. 2. In terms of the mpact of the Crss as captured by the varous annual dummes there s a negatve sgn on the general dummes for ASEAN for and , although they are not sgnfcant. The mpact of the Crss of 1997 on performance n and relatve to PRC s n prncple ambguous snce whlst supplysde dsruptons wll reduce compettveness the exchange rate deprecatons wll ncrease t. Sgnfcant negatve annual dummes for are found for Indonesa, the Phlppnes and Sngapore, and for Indonesa and the Phlppnes n and The apparently unusual case s Malaysa where the dummy for s postve and sgnfcant, mplyng that ceters parbus there are specal characterstcs of , whch rase the compettveness of the economy relatve to PRC n the US. However, ths result s consstent wth the facts, snce for total exports to the US the gap between Chnese export growth and Malaysan export growth for (at 8.7% s lower than that for the perod -97 (at 14.5%. 3. Some dfferental patterns n terms of loss of compettveness are found between countres (see table 5 for a summary. Sngapore has sgnfcant losses that are greater n ts more specalzed actvtes n eght out of the nne categores, wth the only excepton beng the small category for automotve products. The Phlppnes, Malaysa and Thaland have sgnfcant losses that are greater n more specalzed actvtes n seven out of the nne categores. Indonesa has sgnfcant losses n fve out of the nne categores and Thaland n four out of nne. No country shows sgnfcant losses n the small category of automotve products. All countres show sgnfcant losses n resource-based manufactures, engneerng, and other low technology goods. Thaland s the man excepton by falng to show sgnfcant losses for prmary products, textles and garments, and electroncs and electrcals. Indonesa s the excepton n the case of other hgh technology products. In all but one case, whenever the wan ndex of technologcal sophstcaton s sgnfcant t has a postve sgn, mplyng that wthn a gven product category losses are lower ceters parbus the more technologcally sophstcated s the product. The excepton s the case of prmary products from Indonesa, however t s unclear whether ths partcular ndex s meanngful for such goods. No ndvdual country has a sgnfcant tendency for an unambguous gan of compettveness vz-a vz PRC n any category. However n a small number of cases there s evdence of a non-zero crossover rate for the RCA at whch compettveness moves from postve to negatve. Ths occurs n the case of prmary products and engneerng n Indonesa. The crossover RCAs above whch there s a loss of compettveness and below whch a gan are 1.24 and 0.76, respectvely. The other case s for engneerng n the Phlppnes where below a relatvely low crossover rate of 0.23 there s a tendency to a gan n compettveness. 9 The hypothess that there s a dfferent pattern n trends of compettveness between ndvdual ASEAN countres and PRC can be sad to be supported, although only weakly gven the broad smlarty n table The most consstent pattern n our results s the fndng that loss of market share s systematcally and negatvely related to the degree of specalzaton n ASEAN and n ndvdual economes relatve to PRC. Ths follows snce whenever the RCA term s sgnfcant (and t s sgnfcant n 36 out of 54 possble cases t has a negatve sgn. Ths means that n terms of fgure 1 the relatve MC curve for ASEAN relatve to PRC 9 These crossover RCAs requre all three terms the constant for each product category, the RCA and the wan ndex - to be sgnfcant. Relatvely few products are lkely to be below the crossover rate of

17 behaves as n the shft from MC to MC, wth the largest relatve rse n the hgher RCA products. In explanng ths shft we can of ths as a favorable relatve shft for PRC, whch converts nto an unfavorable relatve shft for ASEAN compettors. FDInduced rsng captal productvty s a smple explanaton. If ASEAN economes have specalzed relatve to PRC n captal ntensve goods any favorable shock for PRC that affects ts captal rental-wage rato can be expected to shft the relatve MC curve upwards n ths way, creatng a loss of compettveness for ASEAN that s greater n ts more specalzed markets. Ths, together we some catch up effect n favor of PRC prncpally for commodtes wth hgh RCAs for ASEAN, appears to be what we have dentfed. Table 5 summarzes the stuaton by country n the dfferent product categores, where we fnd statstcally sgnfcant results. Table 5 US market: summary. Sector/country ASEAN Indones a Phlppnes Malaysa Thaland Sngapore Prmary product, - Resourcebased, + manufactures Textle, garment and - footwear Automotve Engneerng, +, +, +, +, + Electronc and electrcal, +, + Other low tech, +, +, +, + Process Other hgh tech Number of categores wth sgnfcant loss a Notes: ndcates compettveness sgnfcantly related to RCA wth negatve sgn + ndcates wan ndex sgnfcant wth postve sgn, - ndcates wan ndex sgnfcant wth negatve sgn. a sgnfcant loss refers to categores where compettveness sgnfcantly related to RCA wth negatve sgn. Japanese market When the same analyss s repeated for the Japanese market (see tables 6 and 7 a few dfferences from the US market can be noted, however n general there s a broad smlarty. Wth only a very few exceptons when a varable s sgnfcant n one market, ts sgn s the same n the other market, even f t s nsgnfcant. The man results from the analyss of compettveness trends n the Japanese market can be summarzed as follows; 17

18 1. For ASEAN as a whole there s evdence of a statstcally sgnfcant loss of compettveness n seven of the nne product categores, as compared wth fve n the US. These are the categores of prmary products, resource-based manufactures, engneerng products, electroncs and electrcals, the other low technology category, process products and other hgh technology goods. The only categores where there s no sgnfcant loss of compettveness are textles and garments and automotve products. As n the US the loss s sgnfcantly greater the hgher s the degree of specalzaton as measured by the RCA, so that for these seven categores losses of market share relatve to PRC are found consstently n the more specalzed areas. Ths appears to be a general pattern snce for ASEAN as a whole the sgn on the RCA term s normally negatve, even when t s nsgnfcant. The only excepton s the case of textles and garments, where t s postve but nsgnfcant. Technologcal sophstcaton, as measured by the wan ndex, s sgnfcant wth a postve sgn for the categores resource-based manufactures and other low technology goods, mplyng that losses are less n the more technologcally advanced products. As n the US there s no systematc tendency for gan of compettveness for ASEAN as a whole n any of the product categores. In Japan, as compared wth the US, there s an even stronger tendency to loss of market share relatve to PRC, that s wdely spread across most products categores, the most mportant excepton beng textles and garments, but here whlst ASEAN as whole shows no sgnfcant tendency to loss of market share three countres (Indonesa, the Phlppnes and Thaland do. 2. In terms of the mpact of the Crss as captured by the varous annual dummes there does appear to be evdence of dfferences between trends n the Japanese and US markets. In Japan for ASEAN as a whole the dummy for s postve although nsgnfcant, whlst that for s postve and sgnfcant. Further and unexpectedly sgnfcant postve annual dummes are found for Indonesa and Thaland n both and The explanaton for Indonesa s better performance n these years relatve to what would otherwse be expected appears to be due to the behavor of prmary exports, whch may have been dverted from the US to the Japanese market n these years. When prmary exports are excluded from the analyss the sgn on the coeffcent for changes to negatve although t s nsgnfcant, and the sgnfcance of the remanng postve coeffcent for dsappears. Thaland s postve term for the annual dummy n reflects the fact that the gap between the growth of Chnese exports to Japan and those of Tha exports n (at 3.1% narrowed compared wth the average for -97 (at 11.4%. Smlarly the postve dummy for agan reflects a narrowng of the gap (at 7.2 % for compared wth 11.4%. 3. Some dfferences n terms of loss of compettveness are found between countres (see table 7 for a summary. Indonesa and Thaland have sgnfcant losses that are greater n more specalzed actvtes n eght out of the nne categores, wth as n the US the only excepton beng the small category for automotve products. Sngapore has sgnfcant losses that are greater n more specalzed actvtes n seven out of the nne categores, the Phlppnes has sgnfcant losses n sx out of the nne categores, and n Japan, Malaysa s the economy wth the lowest number wth four out of nne. As n the US, no country shows sgnfcant losses n the small category of automotve products. All countres show sgnfcant losses n engneerng, electroncs and electrcals and other hgh technology goods. The Phlppnes s the excepton n falng to show sgnfcant losses for prmary products, the Phlppnes and Malaysa do not show sgnfcant losses for resource base manufactures and Malaysa and Sngapore are the exceptons for textles and garments. Malaysa s the only excepton n the case of process products. and electroncs and electrcals. Indonesa s the excepton n the case of other hgh technology products. As n the US whenever the RCA term s sgnfcant ts sgn s always negatve mplyng that losses are greater n more specalzed actvtes. In the Japanese market the RCA term s 18

19 negatve and sgnfcant n an even hgher proporton of the possble cases (40 out of 54, as opposed to 36 out of 54 n the US. 4. No ndvdual country has a sgnfcant an unambguous tendency for a gan of compettveness vz-a vz PRC n any category. However, n Japan we fnd a few more plausble cases of a sgnfcant non-zero crossover RCA. For ASEAN as a whole the crossover RCAs, below whch there s a gan of compettveness are 0.33 for resource-based manufactures and 0.94 for other hgh technology products. For ndvdual countres there are crossover rates of 0.65 for resource-based manufactures and 2.0 for other low technology products n Indonesa, and of 0.32 and 0.58 for engneerng and other low technology products, respectvely for the Phlppnes. These are exceptons however and for the bulk of categores and countres the predomnant result s a loss of compettveness. The mpact of the wan ndex s the man factor n explanng these non-zero crossover rates. In all cases, whenever the wan ndex of technologcal sophstcaton s sgnfcant, t has a postve sgn, mplyng that wthn a gven product category losses are lower ceters parbus the more technologcally sophstcated s the product. Smlar results for the wan ndex between the US and Japanese markets are found for engneerng (where t s postve and sgnfcant for the Phlppnes and Sngapore n both markets and other low technology products (where t s postve and sgnfcant for ASEAN as a whole and for Thaland and Sngapore n both markets. 19

20 20 Table 6 Japanese Market: Results for ASEAN and ndvdual countres. Country ASEAN Indonesa Phlppnes Malaysa Thaland Sngapore I ( ( ( ( ( (0.7 II -0.6 ( ( ( ( ( (1.1 III 0.04 ( ( ( ( ( (1.2 IV -1.4 ( ( ( ( ( (4.8 V -0.2 ( ( ( ( ( (0.4 VI -0.2 ( ( ( ( ( (0.4 VII -0.6 ( ( ( ( ( (1.4 VIII -0.2 ( ( ( ( ( (2.3 Constant IX -2.3 ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( (0.04 I ( ( ( (0.04 II ( ( ( (0.02 III ( ( ( (0.04 IV ( ( ( ( ( (0.1 V ( ( ( (0.02 VI ( ( VII ( ( ( (0.04 VIII ( ( ( ( ( (0.03 IX -0.1 ( ( ( ( ( (0.04 Annual dummy ( ( (0.02 RCA term Annual dummy ( (