New Rates from New Weights

Transcription

1 WP/05/99 New Rates from Tamm Bayoum, Jaewoo Lee, and Sarma Jayanth

2 2005 Internatonal Monetary Fund WP/05/99 IMF Workng Paper Research Department New Rates from Prepared by Tamm Bayoum, Jaewoo Lee, and Sarma Jayanth 1 Authorzed for dstrbuton by Gan Mara Mles-Ferrett May 2005 Abstract Ths Workng Paper should not be reported as representng the vews of the IMF. The vews expressed n ths Workng Paper are those of the author(s) and do not necessarly represent those of the IMF or IMF polcy. Workng Papers descrbe research n progress by the author(s) and are publshed to elct comments and to further debate. Ths paper descrbes the result and the methodology of updatng nomnal and real effectve exchange rate weghts on the bass of trade data over The underlyng framework s an updated verson of the IMF s current effectve exchange rate calculaton, whch uses weghts largely based on data. Snce then, substantal changes have occurred n nternatonal trade relatons, warrantng a recalculaton of effectve exchange rate ndces on the bass of new trade patterns. Updated weghts show that the Unted States and developng countres (most notably Chna) have grown n ther mportance n global trade, whle Japan and the European Unon have declned, wth substantal mplcatons for the path of the dollar and exchange rate effects of emergng market crses snce JEL Classfcaton Numbers: F10, F30 Keywords: Weght, effectve exchange rate Author(s) E-Mal Address: tbayoum@mf.org; jlee3@mf.org; sjayanth@mf.org 1 We owe many thanks to Ercument Tulun for excellent assstance n complng the trade data, as well as to Teng-Sew Boxall for the generous help wth old weghts and toursm data. We benefted greatly from comments by numerous colleagues n the Fund, ncludng Ketl Hvdng, Gan Mara Mles-Ferrett, and Alessandro Zanello.

3 - 2 - Contents Page I. Introducton... 3 II. Now and Then... 4 III. Deconstructng the Weghts... 5 A. The Old (Exstng) Weghts... 5 B. The... 6 C. Commodty Weghts... 7 D. Manufacturng Weghts... 8 E. Toursm Servces Weghts... 9 IV. Reconstructng the Effectve Exchange Rates... 9 A. General Trends... 9 B. Exchange Rates vs-à-vs Subgroups of Tradng Partners C. Three Exchange Rate Events V. Conclusons Annexes I. Data II. Formula for Weghts III. REER Alegbra IV. Expandng the Calculaton of Weghts References Tables 1. New Trade Weghts Dfference n Trade Weghts Weght of Euro Area by Dfferent Methods Dfference n Commodty Shares Importance of Imports n Overall Manufacturng Weghts Importance of Thrd Country Components n Manufacturng Export Weghts Percent Change n Real Exchange Rates Durng Asan Crss Percent Change n Real Exchange Rates Durng Argentne Crss CPI-Based Real Exchange Rates Durng Unted States Dollar Deprecaton Real Effectve Exchange Rates Durng Unted States Dollar Deprecaton Comparson of Exchange Rate Indexes Fgures 1. CPI-Based Real Effectve Exchange Rate Index Rato of Old Index to New Index Old and New Exchange Rate Index Relatve to Subgroups ULC-Based Real Effectve Rate Index... 53

4 - 3 - I. INTRODUCTION Ths paper updates the weghts for effectve exchange rate calculatons, usng trade data over the perod. The weghts currently used n effectve exchange rate ndces publshed n Internatonal Fnancal Statstcs are based on data, wth an adjustment to ncorporate transton countres a few years later. 2 Naturally, these weghts fal to reflect developments n nternatonal trade relatons durng the subsequent decade, whch was punctuated by rapd globalzaton and rsng mportance of many emergng market countres n the global tradng system. Outdated weghts can lead to an ncorrect assessment of the development n the effectve exchange rate, a key nput for the macroeconomc analyss of open economes. A promnent example can be found n the recent dscusson of the U.S. current account defct and exchange rate. Wth the buldup of the U.S. current account defct to a hstorc level accompaned by a substantal apprecaton of ts real effectve exchange rate a consensus appears to have emerged on the nevtable downward correcton n both the current account defct and the real effectve exchange rate of the Unted States. However, the assessment of necessary correcton n the exchange rate wll vary wth prevalng trade patterns of the Unted States. Another example can be found n the growng mportance of Chna n global trade, whch s begnnng to have wde-rangng economc mplcatons. However, data from the late 19s and early 1990s cannot help us to assess the ground that Chna has ganed, or ts economc sgnfcance. Whle the rse n Chna s role s obvous to any observer, the queston remans whose presence has dmnshed as the mrror mage. To gan nsght on the effect of recent trade patterns on effectve exchange rate calculatons, ths paper updates the weghts usng detaled trade data for 164 countres that account for nearly percent of global trade. 3 The results are dscussed n the followng order. Secton II compares the new and old (exstng) weghts, and provdes an overvew of changes. Secton III dscusses n detal how the calculaton of the new weghts dffers from the exstng calculaton. The basc approach s dentcal to that used n the exstng calculaton, whle some mprovements have been made to the calculaton of the mportance of domestc sales and thrd-market weghts. Secton IV dscusses the approxmate effect of new weghts on the path of selected exchange rates over the 1990s, and Secton V concludes. Annexes contan techncal detals, such as descrptons of the data and the formulae for calculatng weghts. 2 The methodology for the effectve exchange rate calculaton was appled wdely as part of the Informaton Notce System (INS), whch had purported to facltate survellance over exchange rate polces. The survellance purpose has snce been de-emphaszed, but one legacy of the INS has been the methodology of calculatng the effectve exchange rate. 3 Hence, the updatng dscussed n the paper s close to but dfferent from the full updatng of exchange rate weghts that encompasses all 184 countres covered by the INS. See Annex 4 for a complete lst of countres and the methods used to update weghts for all countres.

5 - 4 - II. NOW AND THEN Effectve exchange rate calculatons start wth constructng the weghts to be appled to each trade partner. An overvew of the results of updatng these weghts s presented n two tables. Table 1 reports the new weghts for a wde range of key ndustral and developng countres. The trade weghts are reported wth respect to ndustral countres further dvded nto the Unted States, the euro area, Japan, and other ndustral countres and developng countres whch are reported along geographcal lnes as Afrca, Asa (further subdvded nto Chna, Assocaton of South East Asan Natons (ASEAN) and the rest), Latn Amerca, the Mddle East, and the transton countres of eastern Europe and the former Sovet Unon. Table 2 uses the same format to report dfferences from the exstng weghts that cover all countres (n the current weght calculaton, there also exsts a dfferent set of weghts calculated only across ndustral countres). The results ndcate several trends n the patterns of trade. Frst, ndustral countres reman at the heart of the nternatonal tradng system, but ther mportance has declned sgnfcantly compared to the prevous exercse. Industral countres stll account for more than half of the trade-based exchange rate weghts for most countres that are reported (Table 1). In some non-ndustral countres that have partcularly close tradng relatonshps wth major ndustral countres (the Czech Republc, Israel, Mexco, and Poland), the weghts of ndustral countres approach or even exceed percent. That sad, the weghts of ndustral countres have almost unversally declned snce the last exercse, often by qute substantal amounts (Table 2). Ths prmarly reflects the globalzaton of world trade. Wthn the ndustral countres, the Unted States and the euro area are domnant, wth the weght of the Unted States generally ncreasng snce the last exercse whle the weght of other areas has declned. The Unted States and the euro area are partcularly mportant for other countres n north Amerca and Europe, respectvely (Table 1). 4 By contrast, Japan s weght s smaller than that of the Unted States n all reported Asan countres, except n Thaland for whch Japan commands a slghtly larger weght than the Unted States. The weght of the Unted States has generally ncreased, most dramatcally for fellow members of the North Amercan Free Trade Agreement (NAFTA), Canada, and Mexco (Table 2). Ths reflects strong growth n the Unted States, possbly aded by the rse n value of the dollar, whch may have affected trade weghts as U.S. producers typcally prce n dollars whle others prce to market n the Unted States. 5 In addton, some of the declne n the weght of the euro area comes from treatng the regon as a sngle bloc rather than twelve dfferent countres. 4 Trade weghts are calculated both for ndvdual euro area countres and for the euro area as one bloc. However, the descrptve dscusson focuses on the euro area as one bloc. 5 Over the perod, U.S. growth n merchandse trade exceeded that of the euro area n both value and volume terms.

6 - 5 - Asa s the most mportant developng country regon, although the mportance of many developng regons s ncreasng over tme (Table 2). Ths reflects the globalzaton of the nternatonal tradng system, as well as the excluson of many transton countres from the last exercse. Emergng Asa (whch excludes Japan) almost unversally has a larger weght than any other developng country regon, the only major excepton beng the mportance of ntra-regonal trade for Latn Amerca (Table 1). There have also been some vsble shfts n the mportance of regons wthn Asa, reflectng growth dfferentals. The ncreased role of Chna snce the last exercse s partcularly strkng, and there has been a generalzed rse n the mportance of ASEAN countres. In contrast, the weghts of other Asa have decreased n many cases, drven by a declne n the relatve economc weght of the newly ndustralzed economes (NIEs comprsng Korea, Hong Kong SAR, Sngapore, and Tawan Provnce of Chna). Nor has there been much ncrease n the weght of Inda. Regonal trade has become more mportant. There s notceable evdence of strengthened regonal tes, reflectng both regonal trade agreements (NAFTA, Sectoral Commsson for the Common Market of the South (MERCOSUR), and the expanson of blateral arrangements wth the European Unon) and the ntegraton of emergng markets nto the global tradng system for example, Asa has become more mportant for Australa and New Zealand. III. DECONSTRUCTING THE WEIGHTS A. The Old (Exstng) Weghts The aggregate trade weghts reflect the sum of weghts on trade n commodtes, manufactures, and servces. The exstng calculatons generate two sets of weghts that dffer n the scope of country coverage, and n whether domestc competton s ncorporated n calculatng manufactures weghts. The frst method to be called the Global System n ths paper covers a large group of (184) countres, and uses only data on trade flows. The famlar CPI-based real effectve exchange rates of the IMF have been calculated by applyng ths Global System. The second method, to be called the Industral System, covers only ndustral countres for whch unt labor costs (ULCs) were avalable but takes account of domestc sales of home-produced goods n every market (see Annex 2 for detals). Ths method has been used to construct the ULC-based real effectve exchange rates of the IMF. Both methods treat dfferent trade categores n a smlar manner. Indvdual commodtes are assumed to be perfect substtutes, so that the assocated weghts depend on the mportance of other countres n the overall supply and demand for a commodty. By contrast, manufactures are assumed to be dfferentated goods so that weghts depend on blateral flows across countres, augmented by the mpact of thrd-market competton n export markets. In the Industral System, these thrd-market effects depend on the mportance of foregn and domestc goods n overall demand, whle the Global System takes a more mechanstc approach of assgnng equal weghts to drect and thrd-market competton. As far as servces are concerned, only trade n toursm s ncluded, and then only for countres for whch toursm s a partcularly mportant part of overall trade. The weghts are calculated n a smlar manner to those for manufactures, usng blateral data on tourst arrvals. These weghts are then combned based on the mportance of dfferent types of trade, so that:

7 - 6 - ( ) ( ) ( ) W = α W M + α W C + α W T j M j C j T j where W ( M), W ( C ), and W ( ) j j j T denote weghts calculated for manufactures, commodtes, and toursm, respectvely between country and j and αm, αc, and α T denote the shares of these three types of trade n the overall trade of country. B. The The new trade weghts ncorporate three major changes to the exstng weghts: A unform methodology s used for 164 countres. The system used for calculatng thrd-market effects n the manufacturng weghts for ndustral countres (Industral System) has been extended to 164 countres, so that the dstncton between the Global System and the Industral System has been abolshed for them. 6 To overcome data lmtatons, several approxmatons are made as dscussed n Annex 1. Servces trade has been ncluded n a more systematc manner. Rather than focusng on toursm, the new weghts nclude all trade n servces n the calculaton. The man ssue here s that no comprehensve data on blateral trade n servces s avalable, except for the blateral trade n toursm that can be proxed by data on tourst arrvals. What work has been done on trade n servces tends to show that t responds to the same basc factors such as dstance, relatve GDP, and cultural lnks that explan trade n manufactures. Accordngly, trade n servces except for toursm s assumed to be dstrbuted n the same manner as trade n manufactures, and the same weghts are used. However, for the countres n whch toursm s a partcularly mportant part of overall trade, separate weghts are calculated for trade n toursm, usng the same methodology used n the exstng weghts (see Annex 2 for detals). Hence, the new weghts are: ( α α ) ( ) α ( ) α ( ) W = + W M + W C + W T j M S j C j T j where W ( M), W ( C ), and W ( ) j j j T are country-weghts for manufactures, commodtes, and toursm; andα M, α S, α C, and αt denote the shares of manufactures, (non-toursm) servces, commodtes, and toursm n overall trade. The sngle euro area ndex s calculated anew. In the exstng weghts, the members of the euro area are counted as ndvdual countres. Whle contnung ths practce for analytc purposes, a sngle ndex s also calculated treatng the euro area as a sngle 6 See Annex 4 for the methods used to calculate trade weghts for the remanng 20 countres.

8 - 7 - entty wth a sngle exchange rate. 7 Indvdual country weghts capture countryspecfc compettveness when nflaton rates dverge among euro countres, and are mantaned as the unt of calculaton for country-specfc polcy analyss. As a supplement, the sngle euro ndex s calculated to assess the euro-area-wde compettveness aganst other major currences, after accountng for ntra-euro-area trade lnkages. Table 3 examnes the mpact of treatng the euro area as a sngle entty. The frst column reports the weghts derved under the new methodology usng ths assumpton, whle the second column shows the weghts that follow when the euro area countres are treated as ndvdual trade enttes and ther weghts are then summed to get a sngle value for the euro area as a whole. To gan some perspectve on the mportance of any dscrepances, the table also reports the values for the euro area by summng exstng weghts for euro-area countres from the Global System. The results ndcate that treatng the euro area a sngle entty tends to reduce ts weght n other countres effectve exchange rates, but that ths effect s a relatvely small part of the overall change between the old and new weghts. For non-ol commodty exporters, however, the aggregaton reduces the weghts of the euro area notceably, as ntra-euro area trade n commodty s netted out. 8 C. Commodty Weghts Commodty trade s assumed to occur n an ntegrated global market, as commodtes are assumed to be perfect substtutes wth a sngle prce (Annex 2 provdes detaled formulae for the constructon of the weghts). As n the earler exercse, commodtes are defned at the 2-dgt SITC level, leadng to 20 dfferent types of commodtes (see Annex 1 for detals). Wthn each commodty category, the weght country assgns to country j s unrelated to blateral commodty trade, but s nstead determned by country j s share n the global market. The overall commodty weght s obtaned by aggregatng ndvdual commodty weghts, wth allowances made both for the mportance of each commodty category n a country s total commodty trade and for the mportance of the country n the global trade of each commodty. Ceters parbus, a commodty category n whch a country commands a more domnant global presence s counted more heavly when ndvdual commodty weghts are aggregated to the overall commodty weght. Trade n petroleum and energy products, however, are excluded from calculaton of commodty weghts, followng the exstng approach to calculatng weghts. Several reasons underle ths choce. Frst, except n the long run, exchange rate changes are not lkely to 7 The euro area s not the only monetary unon n exstence. Whle ths paper chose the euro area as the most conspcuous example n terms of ts global economc weght, smlar calculatons can be made for other monetary unons (e.g., the West Afrcan Economc and Monetary Unon) as needed for polcy analyss. 8 Intra-euro area trade n manufacturers s also netted out from trade statstcs, but reflected n weghts as domestc sales.

9 - 8 - have much effect on trade n ol or gas. Varable costs account for a very small porton of ther producton costs, and thus exchange rate varaton can exert only a lmted effect on producton decsons. Next, the energy sector s largely segmented from the rest of the economy, except for ts contrbuton to the state budget through energy revenues. The eventual effect of the energy sector on the rest of the economy s affected more by the government s spendng decson than by the exchange rate varaton. Fnally, the world ol market s strongly nfluenced by cartels, and exchange rate varatons have only ndrect effects on the market. Table 4 reports the mportance of commodty trade (n overall trade) for a range of ndvdual economes. The hghest shares are for tradtonal non-ol commodty exporters, wth commodty shares exceedng 20 percent for Chle, New Zealand, Argentna, Russa, Australa, and Brazl. At the other end of the scale, n Sngapore and Tawan Provnce of Chna, commodty trade represents about 5 percent of overall external competton. Compared to the exstng Global System, the relatve mportance of (nonol) commodtes n overall trade has declned across the spectrum, partly owng to the ncluson of servces trade under the new system. D. Manufacturng Weghts Unlke commodtes, manufactures are assumed to be dfferentated goods that are mperfectly substtutable across countres (Annex 2 provdes detaled formulae for the constructon of the weghts). The aggregate manufacturng weghts consst of two effects, the competton through mports of manufactures and through exports of such goods, wth the relatve mportance dependng on the relatve sze of these two flows. Wthn exports, the weghts reflect both the drect competton wth the producers n the destnaton country and the (ndrect) competton wth them n thrd-country markets thus called the thrd-market effect. In the new calculatons as n the Industral System the mportance of the thrdmarket effect s determned by the relatve mportance of mports of manufactures versus sales of home products of the destnaton country (hence the weght s smaller the more closed the country). By contrast, the Global System arbtrarly assgns equal weghts to drect and thrd-market competton. Table 5 presents relatve weghts assgned to manufacturng mports under the new system and the old Global System for the set of countres ncluded n Table 1 (the weght for exports s smply one mnus ths value). The countres wth the hghest mport weghts are the commodty exporters such as Australa, Chle, Argentna, and New Zealand, as such countres mport many more manufactures than they export. By contrast, the lowest weghts go to economes wth few natural resources that mport commodtes and export manufactures, such as Hong Kong SAR, Sngapore, Tawan Provnce of Chna, and Japan. The mddle group generally ncludes economes wth more mxed tradng patterns, such as the euro area and the Unted Kngdom. The Unted States has a very hgh weght accorded to mports wthout beng a commodty exporter, reflectng the large underlyng trade defct. The old Industral System weghts show a smlar pattern to the new weghts, whle the old Global System weghts are less easy to nterpret.

10 - 9 - Table 6 presents the relatve mportance of thrd market competton versus blateral export competton n the same format as Table 4. In the new weghts the mportance of thrd-market competton depends on the openness of the countres to whch exports are sent. Hence, thrd market weghts are relatvely small for countres such as Canada and Mexco, whch export manly to the relatvely closed U.S. market, and are larger for countres such as Sngapore, Australa, and Inda, whose man export markets are the relatvely open Asa regon. Notably, all of these weghts are below ½, the value assgned to thrd-market weghts n the exstng Global System. The exstng weghts n the Industral System show a generally smlar pattern to those from the new methodology, wth the excepton of New Zealand, whose thrdmarket weght far exceeds ½. E. Toursm Servces Weghts For countres that are heavly dependent on trade n toursm servces, the toursm weghts are calculated n the same manner as the Industral System for manufactures weghts (detals n Annex 2). Lke manufactures, toursm servces are vewed as dfferentated products, except that the product s sold by brngng toursts nto a country. IV. RECONSTRUCTING THE EFFECTIVE EXCHANGE RATES Ths secton examnes the mplcatons of the new effectve exchange rate weghts for the analyss of exchange rate movements snce 1995, the perod most relevant for polcy analyss and also that for whch the new weghts are most applcable. Gven a set of weghts for country on partner countres ( W j for j ), real effectve exchange rate (REER) ndces are calculated as a geometrc weghted average of blateral real exchange rates between home country and ts trade partners. Specfcally, the REER ndex of country s calculated by E PR = j PR j j W j where j refers to trade partners, P s are CPIs, and R and R j are blateral nomnal exchange rates of country and j aganst the U.S. dollar (measured n U.S. dollar per local currency). A. General Trends Fgures 1 and 2 graph real effectve exchange rate ndces for a wde range of countres snce the start of 1995 (keepng 1995 average equal to ), calculated vs-à-vs about 40 major trader countres. Fgure 1 reports new and exstng effectve exchange rates, as well as natonal estmates for the Unted States, the euro area, and Japan. Fgure 2 presents the same nformaton, but as the rato of the exstng real exchange rate ndces to the newly calculated real exchange rate ndces, to assess the dfference between two real exchange rate ndces from another angle. The most notable change for the major currences s the more muted apprecaton and subsequent deprecaton of the U.S. dollar usng the new weghts. The U.S. real exchange

11 rate based on new weghts rose by some 25 percent between 1995 (as a whole) and February 2002 rather than the 40 percent found usng the exstng weghts, and fell less subsequently. Ths smaller apprecaton s not offset by smaller deprecatons of currences such as the euro or the yen. Rather, there appears to be a tendency for most currences to have a smaller apprecaton or larger deprecaton under the new weghts most real effectve ndces have smaller numercal values. Ths seemngly paradoxcal result reflects underlyng changes n nternatonal trade relatons. The key here s the ncreased weght of the Unted States n most other countres effectve exchange rates, and the rsng mportance of developng countres n the U.S. effectve exchange rate. Between 1995 and early 2002, many countres experenced a sgnfcant blateral real deprecaton aganst the U.S. dollar, whch was only partly reversed subsequently. As the U.S. dollar s generally accorded a hgher weght n the new calculatons, ths means that outsde of the Unted States exchange rates have tended to deprecate more on a multlateral bass. By contrast, the new calculatons for the Unted States put more weght on developng countres, whose exchange rates have changed less aganst the dollar (as shown n the bottom panel of the data for the Unted States). Hence the unntutve result that most real effectve exchange rate ndces are numercally smaller usng the new weghts snce The U.S. real exchange rate ndex calculated under the new weghts have been much closer to the ndex calculated by the U.S. authortes. The U.S. panel of Fgure 1 shows that the real exchange rate ndex based on the new weghts has tracked the Federal Reserve Board (FRB) real exchange rate ndex (whch uses weghts that are updated from year to year) much more closely than the real exchange rate ndex based on the exstng weghts. 10 For the euro and the yen, all three ndces are much closer to one another than they are for the U.S. dollar. Compared perod by perod, the new and old ndces exhbt substantal dfference. Fgure 2 presents the rato of the old real exchange rate ndex to the new ndex, whch measures the dfference between the two ndces n each month. The dfference mostly ranges from 2 to 6 percent of the new ndex, but gets close to 10 percent of the new ndex n several nstances. The U.S. dollar n 2002 at the peak of ts apprecaton s the frst such nstance, and other smlar nstances arse n the currences of Canada, Argentna, and Mexco over the perod. The dfference between the new and old ndces s also conspcuous n many emergng market currences, ncludng those of Brazl, Chna, and Inda. B. Exchange Rates vs-à-vs Subgroups of Tradng Partners The exchange rate ndces can be calculated separately vs-à-vs subgroups comprsng developng and advanced countres, to llustrate the roles of two groups. Fgure 3 presents the exchange rate sub-ndces measurng only the contrbuton of ether ndustral or developng countres (the overall effectve exchange rate s thus a sum of these two ndces). It comes out 9 See Annex 3 for an llustratve algebrac analyss of a three-country example. 10 See Leahy (1998) for a dscusson of the FRB ndex.

12 clearly that exchange rate fluctuatons are larger aganst ndustral countres than aganst ther developng counterparts, partcularly for ndustral countres. Ths s not lmted to the largest traders (such as the Unted States and euro area) aganst whch many countres formally peg, but s also true for the smaller ndustrals. 11 It probably reflects a range of ssues, ncludng the fact that many emergng market countres are more open to trade, have trade patterns whch are often more concentrated and hence dependent on specfc currences, and often borrow nternatonally n the currences of ther major tradng partners. All of these wll create a desre to lmt exchange rate fluctuatons aganst major tradng partners the socalled fear of floatng syndrome (Calvo and Renhart, 2002). In addton, the group comprses a larger number of ndvdual countres so fluctuatons n ndvdual countres may tend to cancel out more. The analyss so far has focused on comparng the REER ndces based on the new weghts wth that from the Global System. We next calculated the exchange rate across only the ndustral countres, n order to compare the real effectve exchange rates based on the new weghts wth the exstng Industral System whch used a more smlar methodologcal approach (the new REERs are calculated usng unt labor costs, as ths s how the real exchange rates are calculated and reported under the exstng Industral System). As can be seen n Fgure 4, the dfferences n the path of the real exchange rates are generally qute small and are largest for Australa and New Zealand, countres where the weght of commodtes n trade has changed sgnfcantly. Ths suggests that the man reason for the dfferences n Fgure 1 are dfferences n methodology and weghts across ndustral and developng countres. C. Three Exchange Rate Events To further llustrate the propertes of the new and exstng weghts, we compare the real exchange rate movements across all countres for three recent epsodes of large exchange rate movements: the Asan crss (June 1997 to January 1998), the Argentne crss (January to September 2002), and the U.S. dollar deprecaton between February 2002 and May Asan crss. Table 7 shows the changes n the two multlateral exchange rates around the Asan crss, from June 1997 to January The deprecatons n the crss countres (Indonesa, Korea, Malaysa, the Phlppnes, and Thaland) are smlar across the two approaches, reflectng the generalzed nature of the fall n ther exchange rates. Elsewhere, exchange rates are generally estmated to have apprecated more (or deprecated less) n real effectve terms under the new weghts than under the old ones. The dfference s partcularly large for economes wth close regonal tes, ncludng Australa, Chna, Japan, New Zealand, and Tawan Provnce of Chna. Ther real effectve exchange rates apprecated by at least 2 percentage ponts more under the new weghts than under the old ones. 11 For example, t s true for Japan despte the fact that many emergng Asan economes are generally consdered to be more concerned wth ther blateral exchange rates aganst the U.S. dollar than the yen, and that the yen-u.s. dollar rate has fluctuated qute sgnfcantly.

13 The Argentne crss. Table 8 compares the changes n the two multlateral exchange rates from January to September of Agan, the mpact on the crss countres (Argentna, Brazl, and the Repúblca Bolvarana de Venezuela) are smlar under the two weghtng schemes. Most other countres are found to have gone through a larger apprecaton or a smaller deprecaton under the new weghts than under the old ones. In partcular, Latn Amercan countres wth close tes to Argentna are found to have experenced smaller real deprecaton under the new weghts, together wth many other emergng markets and the Unted States. The real deprecaton of the Unted States and some closely lnked countres s lkely to have been drven by the trend deprecaton of the dollar that started n February Currences of other ndustral countres, whch generally apprecated durng the crss, are found to have apprecated by more under the new weghts. Deprecaton n the U.S. dollar snce early 2002 (Table 9). The U.S. dollar deprecated by about 10 percent from the peak of February 2002 to May 2004 under the new weghts, 4 percentage ponts less than under the exstng weghts for all countres. The smaller dollar deprecaton under the new weghts s agan attrbutable to the ncrease n the mportance of developng countres for the U.S. trade, and to the relatve stablty of the exchange rates between these developng countres and the Unted States. Because of the larger weght of the Unted States n other countres trade, however, other currences are generally found to have apprecated by a larger margn (e.g. euro) or to have deprecated by a smaller margn (many developng countres). The dfference s most notceable for the Western Hemsphere countres, ncludng Canada. These dfferences are much less stark f the comparson s made only wth other ndustral countres, whether the weghts are taken from the Global System or Industral System (Table 10). Gven the sgnfcant dfferences between dfferent exchange rate ndces, Table 11 compares two IMF exchange rate ndces and those constructed by the authortes for the dollar, euro, and yen. For the U.S. dollar, the FRB ndex appears to be much closer to the IMF ndex based on new weghts than the exstng ndex based on old weghts. The contrast s much smaller for the euro and yen. V. CONCLUSIONS Comparng trade weghts based on data 10 years apart, several changes n the global trade pattern stand out. Whle ndustral countres reman the domnant force n the global tradng system, ther relatve mportance has declned, beng replaced by emergng market countres ncludng Chna. In contrast to the relatve declne n the mportance of ndustral countres as a whole, the weght of the Unted States has ncreased for most tradng partners. At the same tme, reflectng the rse n regonalsm, the weghts of regonal tradng partners have ncreased for countres n the NAFTA, Latn Amerca, and (South) East Asa. Applyng new weghts to calculate effectve exchange rates, dfferent pctures emerge for several exchange rate epsodes. Startng n 1995, the new real effectve exchange rate ndex for the U.S. dollar apprecated much less n the lead-up to ts February 2002 peak than the exstng ndex. Subsequently, the new ndex also deprecated less than the exstng ndex. In both cases, the new ndex s found to have moved much more closely to the FRB ndex than the exstng ndex that was calculated on the bass of old trade data.

14 Durng the Asan crss n the late 1990s and the Argentne crss n 2002, the real effectve exchange rates of ndustral countres are found to have apprecated more under the new weghts than under the old weghts. Ths contrast s consstent wth the rse n the mportance of crss countres n world trade over the last decade. Beyond crss perods, the much publczed symptom of fear of floatng s observed n the real exchange rate between ndustral countres and developng countres as a bloc. The real exchange rates of ndustral countres calculated vs-à-vs developng countres look almost constant, relatve to ther real exchange rates calculated vs-à-vs the rest of ndustral countres.

15 ANNEX I Annex 1. Data A summary of our methodology helps to put the data dscusson n context. We separately calculated for each country (normalzed) partner compettveness weghts n three categores of trade, namely, a) commodtes, b) manufactures, and c) toursm. Trade n servces, other than toursm, was assumed to follow a pattern smlar to trade n manufactures, and no separate weghts were calculated for ths category of trade. The three sets of partner weghts were then aggregated to obtan an overall set of compettveness weghts agan, for each country by weghtng them by the proporton of trade n the respectve trade categores. For ths purpose, trade n non-toursm servces was lumped wth trade n manufactures, as both are assumed to behave smlarly. a) Merchandse trade: Data was obtaned from Unted Natons COMTRADE database at the SITC double-dgt level on a blateral bass. Averages over (or as avalable n the perod) were used n the calculatons. Blateral trade flows made t possble to correct for ntra-euro-area trade n constructng euro area seres from ndvdual euro area member country trade flows. Commodty categores were dstngushed, n our exercse, at the SITC double-dgt level. They comprse SITC sngle-dgt codes 0, 1, 2, 4, and SITC 2 dgt code 68 (non-ferrous metals). (See Table A.1 for all correspondng double-dgt codes and category descrpton). Trade n each commodty category s assumed not to be dstngushed by source (.e., mports of the same commodty from dfferent countres are perfectly substtutable). Under ths assumpton, only total trade of each country by commodty group s needed to calculate the compettveness weght to be accorded to the country ether as a compettve producer or a consumer of that commodty. Blateral trade from COMTRADE was aggregated by commodty for each country to create the seres needed. Weghts were then calculated for each commodty and then aggregated nto one overall commodty weght 12 usng proportons of trade n the varous commodtes for each country. All other SITC codes, except the fuels group (sngle-dgt code 3), were aggregated nto a sngle manufactures group. 13 Fuels were thus excluded from the exercse. The manufactures group, n contrast to the commodtes group, s just a sngle composte group, trade n whch 12 The euro area fgures less promnently as a commodty compettor n our calculatons, because ntra-euro-area commodty trade flows are no longer ncluded n total commodty trade. An approach, of the knd employed for manufactures, that also takes nto account domestc demand would correct for ths problem, but domestc demand would be hard to comple across a large set of countres for each commodty at the SITC double-dgt level. 13 Hong Kong SAR s mports were adjusted for re-exports. Imports, as obtaned from COMTRADE, were nearly an order of magntude larger than exports and clearly seemed to nclude huge amount of merchandse reexported va Hong Kong SAR. Therefore, Hong Kong SAR s mports were corrected for re-exports, assumng a margn of 15 percent. The reexports seres was also obtaned from COMTRADE.

16 ANNEX I s dstngushed by source. Calculaton of manufacturng weghts, therefore, requres blateral detal. For many countres wth two observatons of blateral trade flow (export from country A to B and mport by country B from country A), the average of the two observatons was used. For countres wthout ther own data on blateral trade, blateral trade data as reported by partner countres were used. b) Servces trade: Data were obtaned from the IMF s World Economc Outlook. Ths was used only to derve the share of manufacturng (.e., manufacturng plus non-toursm servces) n total trade. c) Domestc sales of (home-produced) manufactures: These data were constructed for each country by subtractng the country s manufactures exports from an estmate of ts U.S. dollar nomnal gross manufacturng output. Gross manufacturng output was obtaned for ndustral countres from the STAN database of the Organzaton for Economc Cooperaton and Development. However, t s not readly avalable from a common source for developng countres. It was, therefore, estmated from net (value added) manufacturng data, whch are avalable from the World Bank for a large number of countres. Based on the observed gross/net rato for ndustral countres, a rato of 10/3 was appled to estmate gross output from the reported net manufacturng output. For two economes Hong Kong SAR and Sngapore the estmaton based on a 10/3 rato produced mplausble results. The value of ther manufacturng exports exceeded ther gross output, a physcal mpossblty, presumably reflectng ther role as a reprocessng base and the host for entrepôt trade. We, therefore, appled a rato of 6/1 for these two economes, whch would then mply a measure of openness (as measured by exports/gross output) that s consstent wth what s observed for smlarly open countres, such as Malaysa, Hungary, and the Czech Republc (see the fgure below). Manufactures 000 Sngapore 000 Malaysa 000 Exports = Gross Output Exports Hong Kong SAR Hungary Czech Rep Gross Output

17 ANNEX I d) Toursm trade: Data on tourst arrvals by country and total toursm exports were obtaned from the World Toursm Organzaton. Toursm data were only kept for countres where toursm exports exceeded a threshold of 20 percent of total exports; for other countres, toursm was consdered not sgnfcant (n lne wth the current approach to weght calculaton), and therefore dropped. Toursm exports of 29 countres, whch met the threshold crteron, were allocated to partner countres based on the number of tourst arrvals from those partner countres. Ths blateral toursm data were used n calculatng toursm weghts n a manner smlar to manufactures. The 29 countres wth blateral toursm data are Albana, Antgua and Barbuda, The Bahamas, Belze, Barbados, Comoros, Cyprus, Domnca, Domnca Republc, Egypt, Ertrea, Fj, Georga, Greece, Grenada, Croata, Jamaca, Jordan, St Ktts and Nevs, Lebanon, Malta, St. Luca, St. Vncent and the Grenadnes, Maldves, Maurtus, Samoa, Seychelles, Uganda, and Vanuatu Four of these countres are excluded from the Industral System but ncluded n the Global System. The four countres are Antgua and Barbuda, The Bahamas, St. Vncent and the Grenadnes, and Vanuatu.

18 ANNEX I Table A.1 Two-Dgt SITC Categores Food Lve anmals chefly for food 00 Meat and meat preparatons 01 Dary products and brds' eggs 02 Fsh, crustaceans, mollusks, preparatons thereof 03 Cereals and cereal preparatons 04 Vegetables and frut 05 Sugar, sugar preparatons and honey 06 Feedng stuff for anmals, not ncl. unml. cereals 08 Mscel. edble products and preparaton 09 Agrcultural Raw Materals Tobacco and tobacco manufactures 12 Hdes, skns and furskns, raw 21 Crude rubber (ncludng synthetc and reclamed) 23 Cork and wood 24 Pulp and waste paper 25 Textle fbers (except wool tops) and ther wastes 26 Ols Ol seeds and oleagnous frut 22 Crude anmal and vegetable materals, n.e.s. 29 Anmal ols and fats 41 Fxed vegetable ols and fats 42 Anmal-vegetable ols-fats, processed and waxes 43 Industral Materals Crude fertlzers and crude materals (excl. coal) 27 Metallferous ores and metal scrap 28 Non-ferrous metals 68 Beverages Coffee, tea, cocoa, spces, manufactures thereof 07 Beverages 11

19 ANNEX II Annex 2. Formula for Weghts 15 Commodty Weghts The weghts for commodty trade are calculated on the assumpton that prmary commodtes are homogeneous goods, thus traded n an ntegrated global market. Wth no dfferentaton among countres n the global market, the weght assgned to a country reflects the mportance of the country n the global market ether as a buyer or as a seller. Let county j s export of commodty h and h M j country j s share of global trade n commodty h. Let s trade n all commodtes. s M + X h h h j j j = h h M n n + X n n be country j s mport of commodty h. Let X h j be h s j be h w be commodty h s share of country w M + X h h h = h h M h + X h The weght that country attaches to country j for trade n commodtes s h h ( ) ws h j Wj P = h h w ( 1 s ) h. The numerator adds up the global mportance of country j n dfferent commodty trade (h), applyng weghts that capture the nternal mportance of each commodty for the trade of country. For example, f country has no trade n certan commodty (k), we wll have w equal to zero and the global share of country j n the trade of commodty k would not enter the weght that country attaches to country j for trade n commodtes. Manufacturng Weghts When Data on Home Sales Are Avalable Two countres and j compete wth each other n every separate market, and accordngly, the weghtng scheme counts dfferent markets separately. Under the more comprehensve method used n the INS Industral system sales data comprse both external trade and domestc sales by home frms. The weght that country assgns to country j depends on the presence of country j n a typcal market k, and the mportance of ths market k to country. The theoretcal bass for ths weght traces back to Armngton (1969) and McGurk (1987). Under assumptons of constant elastcty of substtuton among products dfferentated along k 15 Ths annex draws heavly on Zanello and Desruelle (1997).

20 ANNEX II natonal borders, t s shown that the compettveness effect of country j s prce movement on country s proportonal to the weghted sum of country j s market share n all markets n whch countres and j meet, wth the weght of each market determned by the mportance of the market to country. The market shares used n ths calculaton are calculated on the bass of both external trade and domestc sales of each country. Algebracally, consder market k n whch producers of country and j compete wth each other. Let X denote the sales of country j n market k. Let s denote country j s market k j share n market k, and sales (output). k w denote the share of country s sales n market k n country s total k j s k j X = k j k X l l w k X = k n X n Then, the weght that country attaches to country j s W ws k k k j j ( M) = k k w k ( 1 s ) Ths measure captures the nteracton between two countres that arses n each market. The degree of competton between producers of country and j n market k s measured by the product of country k s mportance for country ( w ) and country j s strength n market k ( s k j ). Ths formula for weght can be decomposed nto three components, whch lends tself to a more famlar nterpretaton. W (, 1 s ) ws + w s + w s j j k k j j k j j j ( M) = k k w k The frst term n the numerator gauges the mport competton, namely the competton between countres and j n the home market of country (market ) that s nduced by country s mports from country j. The second term gauges the drect export competton, namely the competton between countres and j n market j the home market of country j that s caused by the export from country to country j. The thrd term gauges the thrd-market export competton, capturng the competton between countres and j n all other markets. k

21 ANNEX II Wth judcous rescalng, ths can be also rewrtten n terms of three self-contaned weghts as follows. (The second term XW s further decomposed nto two n a later equaton.) IM X W ( M) = λ MW + λ XW, j j j j where MW refers to the normalzed blateral mport weght, XW refers to the normalzed export weght that ncorporates both drect and thrd-market compettons, and λ s denote the weghts assgned to each component n constructng the overall weght W. In terms of market shares( k k s j ) and sales shares ( w ), these terms can be wrtten as follows. MW j s j = s l l XW λ λ ws k k k j j = k k w k w ( 1 s ) ( 1 s) ( 1 s ) M = k k k w w k ( 1 s ) ( 1 s ) k X k = k k w k The mport weght ( MW ) s country j s share of country s mports, and the export weght ( XW ) can be nterpreted as the rato of the ntensty of competton between countres and j to the ntensty of competton between country and all other compettor countres. The coeffcent on the mport weght ( λ ) gauges the relatve mportance of the domestc M market n terms of country s competton wth all other countres algebracally, the rato of the measure of the domestc-market (market ) competton between country and all others to the sum over all markets of the measures of competton between country and all others. X The coeffcent on the export weght ( λ ) gauges the relatve mportance of all export markets n country s competton wth all other countres. The export weght can be further decomposed nto the two components: XW = µ BXW + µ TXW, BX TX j j j where BXW denotes the normalzed blateral export weght, TXW denotes the normalzed export weght for thrd-market effects, and µ s denote the weghts assgned to each. Wrtten n terms of market and sales shares, these terms are:

22 ANNEX II BXW TXW ws j j j j = k k k ws k ws k k k, j j j = k k k w k s s k k k BX k k µ = k k w k µ ( 1 ) ws ( 1 s ) ( 1 ) ( 1 s ) w s s k k k TX k k = k k k w BX The coeffcent on the blateral export weght ( µ ) gauges the mportance country s competton wth home producer n each export market country k n market k relatve to country s competton wth all others n each export market. The coeffcent on the TX thrd-market export weght ( µ ) gauges the relatve mportance country s competton wth thrd-partes those other than countres and j n each export market. When Data on Home Sales Are Not Avalable The other method that underles the Global System uses market share measures that are calculated on the bass of exports only, excludng home-market sales by domestc producers. The formula contans three terms, each correspondng to mport competton, drect export competton, and ndrect (thrd-market) export competton. These three terms are conceptually smlar to the three same-named terms under the Industral System, but algebracally dffer n that they do not nclude home-market sales. W M s w where ws k k M X 1 j 1 k, j j j ( ) = β j + β + k k 2 2 w 1 k ( s ), s k j k X j = : country j s share n all manufacturng mports of country k, X l k k l k k X w = : the share of exports to country k n all manufacturng exports of country, X n n

23 ANNEX II β β X M l l = n X l l + X n X n X n = n X l l + X n : share of mports n country s trade n manufactures, and : share of exports n country s trade n manufactures. The most conspcuous contrast between the two methods, noted n Table 6, les wth the relatve weghts assgned to drect export competton and thrd-market export competton. The method of the Global System assgns constant and dentcal weghts to both effects, n contrast to the more comprehensve method of the Industral System whch allows relatve weghts to be determned by trade patterns. Toursm Servces Weghts The weghts are smlar to the manufactures weghts under the INS Global System, whch excludes home sales by domestc producers from calculaton. k k 1 1 w ( ) ( ), T sj T M X j k j Wj ( T) = β ( T) sj ( T) + β ( T) w ( T) + k k 2 2 w ( T)(1 s ( T)) k In the above formula, coeffcents β s denote the share of mports and exports n country s k trade n toursm servces, s ( T ) denotes country j s share of all expendture on foregn j k toursm spent (mported) by resdents of country k, and w ( T ) denotes country k s share of all foregn toursm servces receved (exported) by resdents n country.

24 ANNEX III Annex 3. REER Algebra We develop a smple example that shows that new trade weghts can numercally ncrease or decrease all multlateral exchange rates n the same drecton. Consder a three-country world. Let matrx A denote the trade weght and vector b denote the log changes n blateral exchange rates aganst the thrd country (denoted as the prce of the thrd currency n terms of the frst and second currences). A 0 = b 1 b b 0 2 = The multlateral exchange rates of all three countres are defned by the followng vector R, where I denotes an dentty matrx. 0 0 ( ) R = A I b Next consder the followng new trade weghts and the assocated real exchange rates. Under the new trade weghts, the mportance of the thrd country rose to an extreme the frst and second countres trade only wth the thrd country A1 = x 1 x ( ) R = A I b Now consder the dfference between the two real exchange rates based on trade weghts A1 and A b 1 R1 R0 = 2 b ( x 2 )( b b ) 2 1