Determinants of Business Cycle Synchronization in ASEAN-5 Countries

Transcription

1 Determnants of Busness Cycle Synchronzaton n ASEAN-5 Countres Thanh To-Trung * Abstract: After severely ht by the Asan crss n 1997, the nterest of formng a common currency area n ASEAN countres has been more urgent to replace current exchange rate regmes that showed sgnfcant flaws durng the crss. Accordng to theory of optmal currency area (OCA), the most mportant requrement for the launch of such an area s the level of busness cycle synchronzaton. In order to now how ASEAN-5 countres are becomng better prepared to form a currency area, ths study ams at nvestgatng the determnants of cycle co-movement n the regon. Based on varous estmaton strateges appled n the lterature such as OLS, Instrumental Varables (IV) regresson, panel technque and panel-based IV approach, the study found that busness cycles n ASEAN-5 are lely to be more harmonzed f more trade ntegraton, especally more ntra-ndustry trade ntensty, more smlarty of ndustral structure and more fscal polcy convergence occur. However, more nter-ndustry trade ntensty seems to de-synchronze the cycles due to the domnance of ndustry-specfc shocs n the regon. The study also shows that the effect of ntrandustry trade s largest among determnants. Fscal polcy convergence s at least as mportant as blateral trade tes n shapng the busness cycle co-movement. These fndngs propose an mportant mplcaton. Accordngly, after the launch of a currency unon, f the ncrease of trade s manly created by nter-ndustry trade, the busness cycle synchronzaton could be weaened and then mae the jonng of unon more costly. Thus the establshment of a currency unon n ASEAN-5 should be ntated for countres wth hgh level of ntra-ndustry trade lns and let the remanng have adjustment tme to enhance trade wthn the same ndustres. * PhD student, Department of Economcs, Unversty of Brmngham, the UK. Emal: tothanh77@hotmal.com 1

2 1. Introducton The Assocaton of Southeast Asan Natons (ASEAN) was establshed n Bango n 1967 wth orgnal fve members: Indonesa, Malaysa, Phlppnes, Sngapore and Thaland or ASEAN-5. It was then joned by Brune Darussalam (1984), Vetnam (1995), Laos and Myanmar (1997) and Camboda (1999). The assocaton has been emerged as one of the regons of most successful economc development n the world. However, hstorcally the fluctuaton of blateral exchange rate has nfluenced heavly the compettveness of ASEAN exports gven common export marets for these countres. When the trade ntegraton has been more ntense recently, the pressure to stmulate the exchange rate stablty among countres has been ncreasng. The creaton of a common currency area n ASEAN seems to be the best opton to stablze the regon s exports and motvate the benefts of trade ntegraton. After severely ht by the Asan monetary and fnancal crss n 1997, the nterest of formng a common currency area has been more urgent to replace current exchange rate regmes that showed sgnfcant flaws durng the crss. In addton, the successful ntroducton of the Euro n 1999 and ts clamed benefts are also encouragng a common currency arrangement n the regon. In realty, some movements have been mplemented to promote regonal fnancal stablty, monetary polcy cooperaton, economc ntegraton and then common currency area. They can be named such as the sgnng of AFTA (ASEAN Free Trade Area) n 1992, the establshment of Manla Framewor Group n 1997, the adopton of ASEAN Vson 2020 n 1997 where a tmetable s set up to create an ASEAN Economc Regon, ASEAN Survellance Process n 1998, Chang Ma Intatve n 2000, etc... Accordng to tradtonal theory of optmal currency area (OCA) ntated by Mundell (1961) and McKnnon (1963), ey crtera to jon a common currency area are trade ntensty and symmetry of busness cycles. That s because the declne n transacton costs assocated wth the use of a common currency has a larger effect on the volume of trade and nvestment flows across countres n the area. In addton, busness cycle synchronzaton plays an mportant role n determnng the costs f scarfyng an ndependent monetary polcy. Whle tradtonal OCA lterature consders these crtera as exogenous, recent lterature treats them as endogenous. One of the most nfluental papers on ths topc s by Franel and Rose (1998). Accordng to the authors, two 2

3 requrements for the launch of a common currency are ) the ntensty of trade among potental members and ) the degree of busness cycle co-movement across countres. Gven the fact that trade ntegraton has been pushed up recently n ASEAN countres, the plausble concerns posed are whether busness cycles n the regon are synchronzed and what determne the cycle synchronzaton? Answerng these questons, we can now whether ASEAN countres are becomng better prepared to form a currency area. The frst concern has been addressed n the prevous studywth the concluson that busness cycles n the regon are hghly harmonzed. The second concern s left to ths chapter. Knowng how possble determnants wll shape busness cycle co-movement n the regon s then sgnfcant and helpful to propose valuable mplcatons for the prospect of a common currency unon n ASEAN. Though many emprcal studes on determnants of busness cycle synchronzaton have been mplemented n the lterature, a few of them focus on ASEAN countres. Among them, Cortnhas (2005) analyzes the busness cycle co-movement n ASEAN- 5 countres but just examnes the effect of ntra-ndustry trade ntensty only. In addton, the methodology to control the endogenety of trade and busness cycle synchronzaton s controversal n the lterature. Franel and Rose (1998) advocate Instrumental Varable estmaton (IV) whle Gruben et al (2002) apply Ordnary Least Square (OLS) and Shn and Wang (2004) mplement panel technque. Therefore, a new study nvestgatng more comprehensvely the determnants of busness cycle coherence n ASEAN countres, applyng all possble regresson methodologes s sgnfcant and worthy to get the whole pcture of busness cycle synchronzaton n the regon. That s also the purpose of ths study. In that tone, ths study ams at answerng two research questons. Frstly, what are the major determnants of busness cycle synchronzaton n ASEAN and secondly, how relatvely mportant of each determnant? To address these two questons, based on Stocman (1988) s theoretcal model, we use four possble estmaton approaches (OLS, IV, panel regresson and panel-based IV regresson) appled to the data over the perod for the emprcal nvestgaton. In ths study, we just focus on ASEAN-5 because some mportant data n the remanng countres n the regon s not avalable and trustful. 3

4 The study s began by a bref summary of relevant studes n the lterature, and followed by theoretcal framewor, data and varable descrpton and some more dscussons about emprcal approaches. Next are statstcal descrpton of varables, estmaton results from all approaches and then the robustness analyss. The study ends wth summary of man fndngs and mplcatons. 2. Bref Emprcal Lterature Revew In the lterature, many factors have been so far beleved to drve busness cycle synchronzaton. They can be grouped nto four man determnants,.e. ) trade ntensty, ) fnancal ntegraton, ) smlarty of ndustral structure or specalzaton and v) smlarty of polces or polcy coordnaton, especally fscal polcy smlarty. Among them, the most promnent and also wdely mentoned n almost studes s trade ntensty. The effect of each of these factors found n the lterature wll be presented n detal as follows Trade Intensty The economc theores provde an ambguous correlaton between trade ntensty and busness cycle synchronzaton. Hstorcally, there are two manstreams concernng ths ssue. The frst s European Commsson Vew orgnated from Kenen (1969), whch beleves that more trade ntegraton wll mae asymmetrc shocs occur less frequently among economes wth a large share of ntra-ndustry trade. Because most trade n EU countres s ntra-ndustry trade, trade ntensty can mae these countres be nfluenced smlarly by shocs and then result n busness cycle harmonzaton. In contrast, the second called Krugman Vew developed by Krugman (1991, 1993) and Echengreen (1992) affrms that hgher trade ntensty causes a potental ncrease n asymmetry shocs due to specalzaton. The ndustral specalzaton resultng from trade s also confrmed by standard trade theory when trade s of the Hecscher- Ohln or the Rcardan type. Then busness cycles are less synchronzed f they are ht by ndustry-specfc shocs. In the emprcal lterature, studes have confrmed the fact that trade effect on ncome correlaton s not only dependent on blateral trade ntensty but also on the structure of trade (ntra- and nter-ndustry trade). Franel and Rose (1997, 1998) s study s one of the most sgnfcant and nfluental papers n the area and wdely cted n almost subsequent artcles n the lterature. Based on data collected from 21 OECD countres from , the authors am to 4

5 nvestgate the relatonshp between trade and busness cycle synchronzaton as a crtera for entry to OCA. Ther emprcal study s just nterested n the effect of blateral trade ntensty on ncome correlaton although t mentons the possble dfferent mpacts of nter-ndustry and ntra-ndustry trade on busness cycle synchronzaton. Ther measure for busness cycle synchronzaton s the correlaton of cycle components of real output, ndustral producton, employment and unemployment. Franel and Rose also use two proxes for blateral trade ntensty,.e. ) blateral trade normalzed by total trade and ) blateral trade normalzed by nomnal GDP. The IV regressons are mplemented n ther wor because, accordng to them, countres tend to ln ther currences wth ther larger tradng partners to acheve the exchange rate stablty n nternatonal trade. The monetary coordnaton wth large tradng partners may cause a postve correlaton between trade and busness cycle synchronzaton. Thus a smple OLS regresson may result n a spurous assocaton between trade and busness cycles. Based on gravty trade model, the authors choose three nstrumental varables to fx the spurousness problem, such as ) the dstance, ) dummy varable for common border and ) dummy varable for common language. The study fnds a strong, sgnfcant and postve correlaton between blateral trade ntensty and busness cycle synchronzaton. Although the theory on the relatonshp between blateral trade and ncome correlaton s ambguous, emprcs are rather consstent. Together wth Franel and Rose (1998), majorty of the subsequent researches show that more trade ntensty may cause more hghly synchronzed busness cycles, regardless of dfferent measures of trade ntensty and the model specfcatons. The fndngs from some next studes on OECD countres are qute consstent wth those from Franel and Rose (1998), such as Otto et al (2001), Gruben et al (2002) and Calderon et al (2002). However, some wors estmate somewhat lower effects of trade than Franel and Rose (1998). For example, usng the same 21 OECD countres as Franel and Rose (1998), Gruben et al (2002) fnd a postve correlaton between blateral trade and busness cycle synchronzaton but the trade effect s only half of Franel and Rose s estmate. Some other semnar artcles on developed countres have the same concluson, such as Clar and van Wncoop (2001). Usng data extracted from 9 US census regons and 14 EU countres, they conclude that the larger trade ntensty and to a more extent the hgher level of 5

6 ndustral specalzaton result n much hgher degree of ncome correlaton among US regons than among EU countres. On the other hands, many studes on developed and developng countres also confrm the postve effect of blateral trade. Based on 147 countres over the perod , Calderon et al (2002) fnd that countres havng hgher blateral trade level experence hghly synchronzed busness cycles. Ths fndng s qute robust to varous measures of blateral trade ntensty and de-trendng technques. In addton, the trade effect seems to be hgher for pars of developed countres than both developng pars and developed - developng pars as well. Dfferent from most emprcal studes whch have used cross-country correlaton as proxy for busness cycle synchronzaton, Kose et al (2003) compute the correlaton of ndvdual country s cycle wth the world aggregate. Wth data from 21 developed and 55 developng countres over the perod , the authors see the effect of rsng trade and fnancal ntegraton on nternatonal busness cycle synchronzaton. They fnd the evdence that the global spllover of macroeconomc fluctuatons enhanced by trade ntensty s mostly lmted to ndustral countres, whch s somewhat smlar to Calderon et al (2002) s fndngs. Baxter and Koupartsas (2004), n a study to nvestgate the robust determnants of busness cycle synchronzaton for over 100 countres both developed and developng, nclude not only blateral trade ntensty but also extent of total trade to examne. Blateral trade effect s found to be robustly postve to the excluson as well as ncluson of gravty varables n the model. Ths also mples that the mpact of blateral trade on busness cycle co-movement s ndependent of the mpacts on trade occurrng through gravty varables. Meanwhle, the total trade extent s found to be negatvely and sgnfcantly assocated wth busness cycle synchronzaton. The study, nevertheless, presents that ths varable s actually not robust regardless of ncluson or excluson of gravty varables. There have been some recent studes on East Asan countres n the emprcal lterature, but ther results are not strong and clear compared wth those on developed countres. Choe (2001) examnes the mpact of blateral trade on the co-movement of macroeconomc fluctuatons n 10 East Asan countres and fnds that more synchronzed busness cycles result from deeper trade nterdependence n the regon. The trade dependence effect s, nevertheless, s not unform across all countres. It s 6

7 just qute strong among ASEAN members. Meanwhle, Crosby (2003) s study, wth data from 13 Asa-Pacfc countres from , shows no dscernble and robust assocaton between blateral trade and busness cycle co-movement. Kumaura (2005) apples the basc framewor by Franel and Rose (1998) wth attenton to the structural features n Asa-Pacfc to nvestgate the relaton between blateral trade and ncome correlatons among 13 countres n the regon over the perod The study exhbts a postve assocaton between the two, confrmng the trade channel through whch macroeconomc fluctuatons n countres are more synchronzed. However, trade channel of what countres produce and export rather than the geographcal trade structure s the most mportant determnant of busness cycle synchronzaton. Accordngly, those countres specalzng n the electroncs sector tend to have hgher degree of cycle co-movement. If all above mentoned wors just focus on the effect of blateral trade, some others have tred to test the dfferent possble mpacts of ntra- and nter- ndustry trade on busness cycle correlaton. Varous approaches to dsentangle those two effects have been appled n the emprcal lterature. A drect approach s used n Gruben et al (2002) s paper, where both ntra- and nter- ndustry trade varables are ncluded n the model together wth gravty varables as ndependent varables. Ther result s consstent wth theory when the ntra-ndustry trade largely contrbutes to ncome correlaton. They also fnd a postve mpact of nter-ndustry trade on busness cycle co-movement. The culprt, accordng to them, could be attrbuted to hgh share of ntra-ndustry trade n total trade, thus ndustry-specfc shocs, through specfcaton, do not overweght the common demand shocs or productvty spllovers. Another drect approach s conducted n Krdmuc (2004). Researchng on OECD countres from , Krdmuc augments Franel and Rose s basc emprcal model by ncludng ntra-ndustry trade varable and fnds no clear assocaton between blateral trade ntensty and ncome synchronzaton. The ntra-ndustry trade, nevertheless, s found to be statstcally sgnfcant and postve mpact on macroeconomc co-movement. Also addng ntra-ndustry varable n the model wth blateral trade and polcy coordnaton varables, Shn and Wang (2004) focus on 12 countres n East Asa over the perod of Ther estmaton results exhbt that blateral trade varable appears statstcally sgnfcant just when ntra-ndustry s not ncluded n the model. Otherwse, the sgnfcance of the varable s coeffcent 7

8 drops substantally, mplyng that ntra-ndustry trade s the mportant determnant of busness cycle synchronzaton rather than nter-ndustry trade or total blateral trade tself. In an ndrect approach, Inlaar et al (2005) nclude specfcaton varables, one of whch s share of ntra-ndustry trade to capture the mpact of ntra-ndustry trade on busness cycle synchronzaton for 21 OECD countres durng In ther wor, varables are measured n a dfferent way from pror studes. They use Fsher s z-transformatons of correlaton coeffcents as busness cycle co-movement because they clam that smple blateral correlaton of some real economc actvtes les between -1 and 1, mang the error terms n the regressons not be normally dstrbuted. In addton, blateral trade ntensty s computed by a common varaton of the sx ndvdual trade ntensty measures. Smlar to most emprcal studes, ntrandustry trade may result n ncome co-movement and trade ntensty s found to be postvely correlated wth busness cycle synchronzaton, but at much smaller degree than Franel and Rose (1998) s research. Meanwhle, Cortnhas (2005) just concentrates on the mpact of ntra-ndustry trade on busness cycle synchronzaton n ASEAN countres. Based on annual data over the perod , the author estmates two separate models. The frst s the replcaton of Franel and Rose s basc model and the second s the model where he uses the square of the dfference n the growth rate of detrended real output as the proxy for busness cycle harmonzaton. Both two specfcatons produce the expected sgns for ntra-ndustry trade coeffcents but they appear not very statstcally sgnfcant except when Indonesa s excluded from the regresson. That s because, accordng to the author, Indonesa s the largest and least open economy n the regon. Though most of the studes have exhbted qute smlar results on the mpact of trade ntensty on ncome correlaton, no consensus of methodology has been observed n the emprcal lterature. Some wors mae use of IV regressons to fx the endogenety and measurement error problems, namely Franel and Rose (1997, 1998), Otto et al (2001), Calderon et al (2002), Cortnhas (2005), etc. Some other authors, such as Crosby (2003), Inlaar et al (2005), etc, clam that IV approach s not approprate because the gravty varables used as nstrumental varables may not only have mpact on trade ntensty but also on other varables that determne busness cycle synchronzaton and thus they advocate OLS regressons n the studes. 8

9 Meanwhle, Imbs (1998) and Shn and Wang (2004) beleve that all possble troubles can be fxed by applyng panel regressons wth fxed effects. A more detaled dscusson on dfferent estmaton methodologes wll be presented n the next secton. Fnancal Integraton Most of the emprcal studes confrm the postve effects of fnancal ntegraton on busness cycle synchronzaton. Kose et al (2003) fnd that, together wth trade ntegraton, the fnancal ntegraton enhances the global spllovers of macroeconomc fluctuatons. Imbs (2004) s study shows that economc actvtes n regons havng strong fnancal connectons are more harmonzed. Accordng to the author, due to hgher specalzaton, the postve drect mpact of fnancal ntegraton on busness cycle co-movement outweghs the negatve ndrect mpact. In usng the same measures for fnancal ntegraton as Imbs (2004), Inlaar et al (2005) do not fnd a robust effect on busness cycle synchronzaton. However, the correlaton of changes n the stoc maret s found to be robust and postve assocated wth ncome correlaton. Jansen and Stoman (2004) suggest the same concluson wth the measure of captal flows as fnancal lnage. Otto et al (2001), nevertheless, do not fnd a sgnfcant mpact of fnancal ntegraton n ther semnal artcle. Smlarty of Industral Structure Smlar to effect of ntra-ndustry trade, countres wth greater smlarty of sectoral structure tend to have more-synchronzed busness cycles f the prmary shocs are ndustry-specfc. Clar and van Wncoop (2001), n a study to compare US and EU, fnd that one of the reasons why cycle co-movement n US regons s much hgher than EU countres s a hgher level of sectoral specalzaton n the latter. Imbs (2001, 2004), Calderon et al (2002), Crosby (2003) also confrm the evdence that the smlarty of economc structure between countres can lead to more busness cycle synchronzaton among them. Especally, Imbs (2004) s wor shows very strong specalzaton effect on cycle harmonzaton, and most of ths mpact s not dependent on fnancal or trade polces. Smlarly, Inlaar et al (2005) wth dfferent measures for specalzaton such as ndustral smlarty, export smlarty and share of ntrandustry trade confrm a sgnfcant assocaton between those varables and busness cycle co-movement. However, n some studes, symmetrc ndustral structure does 9

10 not appear sgnfcantly and robustly correlated wth busness cycle synchronzaton, such as those by Otto et al (2001), Baxter and Koupartsas (2004). Fscal Polcy Smlarty The effect of fscal polcy smlarty s theoretcally ambguous. If fscal shocs drve busness cycles, the factor can be assocated wth ncome co-movement. However, f fscal polces are mplemented to respond asymmetrc shocs, the effect may be opposte. Emprcally, Clar and van Wncoop (2001) do not fnd any relatonshp between degree of fscal polcy coordnaton and busness cycle correlaton. Smlarly, Shn and Wang (2004) fnd postve but not sgnfcant effect of fscal polcy smlarty on ncome synchronzaton. However, n Inlaar et al (2005) s study, ths effect s strongly sgnfcant and even as large as the effect of trade ntensty. Darvas et al (2005), based on 21 OECD countres over the perod , conclude that fscal convergence as measured by persstent smlarty of rato of government surplus/defct to output tends to mae busness cycles fluctuate more closely. In bref, trade ntensty has been attracted to many attentons n lterature. Most of emprcal wors fnd postve assocaton between blateral trade and busness cycle synchronzaton. However, the trade effect n developng countres seems to be lower than developed countres. Some other studes nclude ntra-ndustry trade ntensty and exhbt a qute strong mpact of ths factor on busness cycle harmonzaton. However, a few wors nvestgate the effect of nter-ndustry trade drectly n the analyss. The effects of fnancal ntegraton, smlarty of ndustral structure and fscal polcy convergence on cycle coherence are not very strongly consensus across emprcal studes compared to the effect of trade ntensty. The estmaton methodologes are also controversal n the emprcal lterature. There are three man regresson strateges appled so far, such as IV, OLS and panel technques, n whch IV regressons are more wdely used n emprcs. In addton, nearly none of studes nvestgates determnants other than ntra-ndustry trade of busness cycle synchronzaton n ASEAN-5 countres. Ths crtcal gap wll be flled n ths study. More nformaton on the relevant researches can be seen n Table 1 as follows. 10

11 Table 1. Summary of Relevant Studes on Determnants of Busness Cycle Synchronzaton Authors Franel and Rose (1998) Otto et al (2001) Clar and van Wncoop (2001) Calderon et al (2002) Gruben et al (2002) Kose and Y (2002) Countres and Perod 21 OECD countres, OECD countres, US regons and 14 EU countres, varous sample perods 147 developng and developed countres, OECD countres, countres, Busness Cycle Sync. Measures Blateral correlaton of detrended GDP, IP, employment and unemployment Blateral correlaton of GDP growth rate Blateral correlaton of GDP and employment, usng HP detrendng and percentage changes Blateral correlaton of detrended output Blateral correlaton of detrended GDP, IP, employment and unemployment Blateral correlaton of detrended output Methodology Instrumental Varables regresson Instrumental Varables regresson Instrumental Varables regresson Instrumental Varables regresson Ordnary Least Square Instrumental Varables regresson Man Conclusons Blateral trade ntensty has a postve and sgnfcant effect on busness cycle synchronzaton Trade ntensty, FDI mpact postvely busness cycle harmonzaton. Meanwhle, spreads, volatlty of nterest rates and exchange rates and ndustral structure have a negatve assocaton wth ncome correlaton. US regons have more harmonzed busness cycles than EU countres. Trade ntensty has sgnfcant mpact on cycle co-movement whle polcy smlarty does not affect synchronzaton. Blateral trade ntensty and more symmetrc producton structure have postve effect on busness cycle synchronzaton. The effect of trade s smaller when producton structure s more dfferent. The mpact of trade s also smaller n developng countres. Intra-ndustry trade has the postve and sgnfcant effect on busness cycle synchronzaton. He mpact of nter-ndustry trade s not clear Blateral trade ntensty nfluences postvely busness cycle harmonzaton. De Haan et al (2002) Crosby (2003) Kose et al (2003) 18 OECD countres, Asan countres, countres, Blateral correlaton of detrended output Blateral correlaton of detrended output Correlaton of ndvdual country s output and consumpton Ordnary Least Square Ordnary Least Square Instrumental Varables regresson Trade ntensty and exchange rate volatlty have postve effect on ncome correlaton Trade s not strongly assocated wth busness cycle synchronzaton but smlarty of ndustral structure have sgnfcantly postve mpact on ncome correlaton. Trade ntensty wth G7 countres has the postve mpact whle captal account restrctons and terms of trade volatlty have 11

12 Shn and Wang (2004) 12 East Asan countres, Imbs (2004) 24 countres, 1980s- 1990s Baxter and Koupartsas (2004) Camacho et al (2005) Inlaar et al (2005) Cortnhas (2005) Davas et al (2005) Over 100 countres, EU countres, Canada, Japan, Norway and US, OECD countres 5 ASEAN countres, OECD countres, growth wth the world aggregates Blateral correlaton of detrended output Blateral correlaton of detrended output Blateral correlaton of detrended output A combned measure of three measures of synchronzaton Blateral correlaton of detrended output and ndustral producton Squared dfference n growth rate of detrended real output Blateral correlaton of detrended output and unemployment. Panel regresson wth fxed effects Three-Stage Least Square Extreme Bound Analyss (EBA) to test robustness of explanatory varables Instrumental Varables regresson Instrumental Varables regresson and Ordnary Least Square Instrumental Varables regresson Instrumental Varables regresson and Ordnary Least Square negatve effect on cycle comovement Intra- rather than nter-ndustry trade nor the volume of trade tself s the man channel through whch the busness cycles become synchronzed Specalzaton and trade ntensty have opposte effects on busness cycle synchronzaton. Fnancal ntegraton not only affects ncome correlaton but also drectly nfluence specalzaton. Robust varables are trade ntensty. The varables whch are not robust are currency unon, total trade extent, ndustral smlarty, factor ntensty, smlarty of mport and export basets Specalzaton, trade ntensty, fscal polcy, labor productvty has sgnfcant effect on busness cycle synchronzaton. Monetary polcy s not related to ncome correlaton Trade ntensty, specalzaton, smlarty of fscal and monetary polces and common currency have postve and sgnfcant mpact on busness cycle comovement. The effect of ntra-ndustry trade on busness cycle synchronzaton n the regon s not very sgnfcant but t becomes hghly sgnfcant f Indonesa s excluded from analyss. Fscal convergence ( n form of persstently smlar rato of government surplus/defct to GDP) s systematcally assocated wth more harmonzed busness cycles. Kamaura (2005) 13 East Asan countres, Blateral correlaton of detrended output Instrumental Varables regresson and Ordnary Least Squar The prmary determnant of the busness cycle synchronzaton across countres s not the geographcal structure of ther trade but what they produce and export. The ncome harmonzaton s qute hgh among those specalzng n the electroncs ndustry. 12

13 3. Data and Methodology 3.1. Theoretcal Framewor In order to nvestgate determnants of busness cycle synchronzaton n ASEAN-5 countres, we wll follow Stocman (1988) and Franel and Rose (1998) to construct a model, where output correlaton can be attrbuted to ndustry-specfc shocs and country-specfc aggregate shocs. The model wll then provde some deas about the channels through whch some factors can have mpacts on busness cycle comovement. The growth rate of real output for home country can be decomposed n the followng equaton: d ln y t = μ d ln y (1) t where d ln yt s the growth rate of real output at tme t, μ s the weght of ndustry n total output ( μ = 1), d ln yt s the growth rate of output n ndustry at tme t. On the other hand, growth rate of output can be expressed as the weghted average of the growth rate of output n dfferent sectors of the economy. We can express equaton (1) n a dfferent way as followng d ln = μ θ + ξ (2) y t t t where θ t s the ndustry-specfc devaton of the growth rate of output n ndustry from the average growth rate of total output at tme t ( ξ t ). Therefore, the growth rate of real output at tme t conssts of weghted average of ndustry-specfc devatons of the growth rate of output n each sector ( rate at tme t ( ξ t ). μ θ ) and the country s average growth Smlarly, the growth rate equaton for foregn country can be presented as t d ln y * t * * * = μ θ t + ξ t (3) where the asters expresses the value for foregn countres. 13

14 Followng Stocman (1988), some followng assumptons are mposed ) Industry-specfc shocs are common across countres. Equvalently, * θ = θ t and they have the same varance t ) { θ } t well. ) { ξ } t v) { θ t } and { t } 2 σ s dstrbuted ndependently of each other across sectors and tme as s dstrbuted ndependently over tme ξ are dstrbuted ndependently of each other. Wth above assumptons, the cross-country covarance of output growth rate s shown as follows * cov ( d ln, d ln y t ) = yt * 2 * μμ σ + cov ( ξ t, ξ t ) (4) The equaton can be rearranged n terms of correlaton of cross-country output growth rate as bellows ρ * = yy * 2 μ μ σ + ξ ρ ξξ * ω y (5) where ρ * s cross-country output growth correlaton, yy μ = μ and then the drect σ σ ςσ * ξ functon of share of ndustry n total output, ωξ y = and ρ * s the correlaton ξξ σ σ y * y y between country-specfc aggregate shocs. So * μ μ and ω ξy are the weghts for the 2 varance of ndustry-specfc shocs ( σ ) and the correlaton of country-specfc aggregate shocs. The equaton (5) mples that the cross-country output growth correlaton or busness cycle synchronzaton depends on the effect extent of ndustry-specfc shocs and country-specfc aggregate shocs. The mpact of some factors on busness cycle synchronzaton can be explaned through these two channels. Followng s the detaled dscussons on ths ssue 14

15 Effect of Trade Intensty Theoretcally, trade ntegraton can have postve mpact on the correlaton of countryspecfc aggregate shocs. There are two man channels. The frst s the spllover of demand aggregate shocs. For nstance, f aggregate demand shocs drve a surge n one country, the effect can spll over to tradng partners due to an ncreased demand for both foregn and domestc goods. The spllover mpact s even larger f t s accomplshed wth more polcy coordnaton resultng from trade ntegraton. The second channel s a more rapd spread of productvty shocs va nward FDI and technology sourcng (as suggested by Lchtenberg and Pottelsberghe (1998)) and through the spllover of nowledge and technology (as suggested by Coe and Helpman (1995)). Therefore, through aggregate demand shoc spllower and productvty dffuson, blateral trade ntensty s postvely correlated wth the second element of the equaton (5) regardless of ts composton (ntra- or nter-ndustry trade). However, the total effect on busness cycle s ambguous because trade components have dfferent possble effects on the frst element of the equaton (5). The mechansm of trade affect on busness cycle synchronzaton wll be clearer f ntra and nter- ndustry trade effects are examned separately. Effect of nter-ndustry trade ntensty Accordng to the standard trade theory, nter-ndustry trade patterns of nternatonal trade s accompaned wth an ncreasng specalzaton n producton resultng from removal of tarff barrers. Therefore, a negatve correlaton between μ and * μ s 2 expected. Gven that σ s always postve, the frst element of the equaton (5) may be negatve. However, as stated before, the effect of nter-ndustry trade on the second element s stll postve. If the ndustry-specfc shocs ( θ t ) domnate common demand shocs and productvty spllovers, nter-ndustry trade s expected to have negatve effect on busness cycle harmonzaton 1. Otherwse, f busness cycles are not domnated by ndustry-specfc shocs, the postve mpact of nter-ndustry trade can be expected. 1 Kaleml-Ozcan et al (2001) suggest another channel to explan the negatve effect of trade. Accordng to them, countres may nsure aganst asymmetrc shocs n face of ntegraton of both fnancal and good marets by dversfy the ownershp, then can afford to specalze the producton structure, creatng better opportuntes for ncome dversfcaton, whch are assocated wth less busness cycle synchronzaton. 15

16 Effect of ntra-ndustry trade ntensty The ntra-ndustry trade does not necessarly cause the specalzaton effect as stated n standard trade theory. Then ndustry-specfc shocs mght affect tradng partners symmetrcally because they are more ntegrated 2. Therefore, n the model, ntrandustry trade does not necessary create negatve correlaton between μ and μ, so the net effect on busness cycle co-movement s expected to be postve 3. In bref, the mpact of blateral trade on ncome correlaton s not only dependent on the structure of trade but also on the relatve mportance of ndustry-specfc shocs and aggregate shocs n drvng busness cycles Effect of Fnancal Integraton The fnancal ntegraton may stmulate producton specalzaton through the channel of captal reallocaton so that t s consstent wth countres comparatve advantages. In the same explanaton manner as the prevous sub-secton, the net effect of fnancal ntegraton on busness cycle synchronzaton may be negatve f the ndustry-specfc ndustry shocs are domnant. However, the fnancal lns can generate a large aggregate demand sde effect and then have postve effect on the second element of the equaton (5), contrbutng to busness cycle co-movement. For example, f customers n other countres mae a sgnfcant nvestment n a partcular stoc maret, a declne n that stoc maret could result n a smultaneous decrease n the demand for nvestment and consumpton n those countres. Moreover, cross-country spllover of busness cycles could be resulted from the contagon effects whch are transmtted va the nternatonal fnancal ln. In addton, n case nternatonal fnancal marets are used to dversfy the consumpton rs, the consumpton across countres can be more strongly correlated. Ths effect s even stronger among developng countres gven ther more volatle output and less dversfcaton of endowment and producton structure. * 2 See Krugman (1993) for more detals. 3 The postve correlaton between trade ntensty and ncome synchronzaton s explaned dfferently n Kose and L (2001). The authors argue that the nternatonal trade observed exhbts ncreasng porton of vertcal or fragmented trade,.e. countres now tend to specalze n partcular stages of a good s producton sequence rather than the entre good. Ths s lely to create more synchronzed busness cycles n response to hgher trade ntegraton. 16

17 Effect of Smlarty of Industral Structure Countres wth smlar ndustral structure or producng smlar type of goods and commodtes seem to have smlar stochastc developments n case of ndustryspecfc shocs (Imbs, 2004). In the model, smlarty of ndustral structure results n a postve correlaton of between μ and * μ and then the frst element of the equaton (5). Further more, as mentoned n Kraay and Venture (2001), the countres wth smlar producton pattern can be more synchronzed even though shocs are purely aggregate regardless of sectors dfferng n ther response to monetary shocs. Therefore, the effect of ths factor s not ambguous. We expect a postve relatonshp between structural smlarty wth busness cycle harmonzaton Effect of Smlarty of Fscal Polcy Theoretcally, the effect of fscal polcy smlarty on ncome correlaton s not clear. Countres may be subject to asymmetrc shocs or symmetrc shocs may have asymmetrc effects due to dfferent propagaton mechansm across countres. If ths asymmetry s persstent and compensated partally by dscretonary fscal polcy, fscal dvergence or fscal polcy dssmlarty can result n busness cycle comovement. In contrast, wthout shocs wth persstent effects or persstent shocs, the factor could have no mpact on busness cycle co-movement. However, fscal polcy, as mentoned n study3, normally pro-cyclcal and can be a source of shocs. So f fscal shocs themselves drve busness cycles, the postve relatonshp between fscal polcy smlarty across countres and busness cycle synchronzaton can be expected Data and Varable Defntons As prevously mentoned, n addton to blateral trade ntensty whch has receved great attenton n both theoretcal and emprcal wors, there are some major determnants of busness cycle synchronzaton such as fnancal ntegraton, specalzaton or structural smlarty and smlarty of fscal polcy. However, blateral fnancal ntegraton s notorously dffcult to measure, even for developed countres (Imbs, 2004). In addton, n case of ASEAN-5, the data on effectve blateral captal flow - one of two man ndces capturng fnancal ntegraton, together wth restrctons on captal flows s not avalable for long tme. Therefore, 17

18 ths factor s excluded, leavng others namely trade ntensty, smlarty of ndustral structure and fscal polcy coordnaton to play n the study. Followng are detals of varable measurement, data sources and some ntal statstc descrptons Busness Cycle Synchronzaton Standard measure for busness cycle synchronzaton n the lterature s the correlaton coeffcent of cyclcal component of economc actvtes (GDP, ndustral output, employment, unemployment, etc ). It s also used n a majorty of emprcal wors such as Franel and Rose (1998), Baxter and Koupatrtsas (2004), Shn and Wang (2004). Followng them, the study s measure for busness cycle co-movement s the blateral correlaton of detrended real output. The technque to solate the cyclcal component of output s HP flter. The source of data s World Development Index (WDI), September 2005, where the data s annually avalable from for ASEAN-5 countres. As mentoned bellows, the data avalablty for other varables s yet shorter, thus the perod for analyss s cut to to acheve data symmetry. In the emprcal lterature, there s no consensus and obvous way to splt the sample perod nto partcular sub-perods to compute the correlaton coeffcents, so we dvde the whole sample nto 7 sub-perods of equal length. The reason s to get suffcent observatons to run regressons gven a short length of data. Snce the sample ncludes 5 countres, the number of observatons s 70 (10 country pars each wth 7 sub-perod observatons) Trade Intensty The study nvestgates not only the effect of blateral trade ntensty but also the effect of ntra-ndustry and nter-ndustry trade on busness cycle synchronzaton. Therefore, 4 measures for ths determnant wll be computed n the study. Blateral Trade Intensty We use four proxes for blateral trade ntensty. The frst varable (BTI 1 ) uses export data only, the second (BTI 2 ) uses mport data only, the thrd (BTI 3 ) uses both export and mport data and the fourth (BTI 4 ) use output data as well. They are defned as follows xj BTI 1 (,j) = x + x j 18

19 mj BTI 2 (,j)= m + m j BTI 3 (,j)= x x j + m + m + x j j + m j x BTI 4 (,j)= y j + m + y j j where x j s total value of nomnal exports from country to country j, mj denotes total value of nomnal mports from country j to country. x s the value of country s exports to all countres. m s the value of country s mports from all countres. y s the output of country. The measures BTI 1 and BTI 2 are smlar to Shn and Wang (2004) whle BTI 3 and BTI 4 are used by Franel and Rose (1998). The export and mport data s collected from World Trade Flows, source compled by Feenstra et al (2005). The output data s extracted from WDI, September Practcally, as suggested by Feenstra (2000), we use only mport data to compute all ndces because reported data are more relable for mports than exports. Accordngly, followng are practcal formulas for blateral trade ntensty: BTI 1 (,j) = x m j + x j mj BTI 2 (,j)= m + m j BTI 3 (,j)= x m + m j + m + x j j + m j m BTI 4 (,j)= y j + m + y j j A hgher value for any above ndex mples more ntense blateral trade across countres. 19

20 Intra-ndustry trade Intra-ndustry trade ntensty s computed usng the tradtonal Grubel-Lloyd (1975) Index. IIT(,j) = 1 - x ( x j j m + m j j ) Or practcally IIT(,j) = 1- m ( m j j m + m j j ) where x j s the nomnal value of exports of product from country to country j and m j s the nomnal value of mports of product from country j to country. The ndex s calculated by usng World Trade Flows, data by Feenstra et al (2005) at four-dgt ndustry classfcatons followng the Standard Internatonal Trade Classfcaton, revson 2. It should be noted that the hgher the ntra-ndustry trade, the lower the second element of the above formula. Ths element s subtracted from 1, mplyng that the ndex IIT wll ncrease when more ntra-ndustry trade occurs. Dsentanglement of blateral trade nto ntra- and nter-ndustry trade Followng suggeston from Gruben et al (2002), we decompose blateral trade ntensty nto nter- and ntra-ndustry trade by usng another two varables: nterndustry trade component (Inter) and ntra-ndustry trade component (Intra). By multplyng ntra-ndustry trade ntensty (IIT) wth blateral trade ntensty (BTI), we can get ntra-ndustry trade component Intra(,j) = IIT(,j). BTI(,j) Thus the nter-ndustry trade component s defned as Inter(,j)=(1-IIT(,j)). BTI(,j) 20

21 Smlarty of Industral Structure The smlarty of ndustral structure s proxed by the correlaton coeffcent of sectoral shares, whch s also computed n Imbs (1998) and Baxter and Koupartsas (2004). SIS(,j) = s s 2 s j s 2 j where s s the output share of ndustry n country. The greater the smlarty n sectoral structure, the hgher the value of SIS. The data source for ths varable s Natonal Accounts Man Aggregate Database, UN Statstcal Yearboo, where GDP s broen down to 7 sectors bac to the year Fscal Polcy Smlarty Followng the suggeston by Shn and Wang (2004), n the study, fscal polcy smlarty s measured by the correlaton coeffcents of the rato of budget/defct to GDP across each par of countres. A hgher value of ths coeffcent n a par of countres mples more smlarty n fscal polcy between those countres. More detals on varable defnton and data sources can be found n Table A.1 (n Appendx) 3.3. Some Econometrcs Issues In the emprcal lterature, some studes advocate IV regressons. Franel and Rose (1997, 1998) argue that countres seem to ln ther monetary polcy and fx exchange rates towards ther mportant tradng partners, ether explctly or mplctly. Then the monetary coordnaton wth large partners may result n a spurous correlaton between trade ntensty and busness cycle co-movement. Some others argue that because trade ntensty s endogenous, the explanatory varables may be correlated wth the error terms, and OLS regressons can produce based and nconsstent estmates. Accordng to IV supporters, all above troubles can be addressed f nstrumental varables for trade ntensty are avalable. These varables should be exogenous determnants of blateral trade but unaffected by other condtons nfluencng busness cycle synchronzaton. Most of the authors choose nstrumental varables provded by the gravty models, such as log of dstance between two 21

22 countres, dummy for geographc adjacency and dummy for common language. These gravty varables regarded as exogenous characterstcs of country pars are beleved to explan great deal of blateral trade. However, some authors crtcze the IV method and support the OLS technque. Gruben el al (2002) clam that IV s not approprate soluton as the gravty varables are lely to have effect on other varables that affect busness cycle coherence. For nstance, countres wth common border are more lely to ln ther monetary polcy or to jon a common currency area than countres wth a long dstance. Furthermore, the mplementaton of sngle currency can stmulate trade ntensty through reducng tradng costs by removng exchange rate rss then the cost of hedgng and decreasng nformaton costs (De Grauwe and Mongell, 2005). Ths relatonshp among busness cycle synchronzaton, trade, gravty varables and other varables can be depcted as the followng fgure: C T V G The regresson equatons correspondng to above fgure are C = α 1 T + α 2 V + ε T = β 1 V + β 1 G + μ (6) V = λg + φ where C s ncome correlaton, T s trade varables, G s gravty varables and V s other varables. Therefore, accordng to Gruben et al (2002), n IV regresson, the nstruments to capture the effects of trade ntensty could ultmately be seen as capturng the effects of last two equatons n (6), then upwardly basng the estmated effect of trade ntensty alone. Meanwhle, some emprcal studes apply panel technque wth fxed effect to examne the determnants of busness cycle synchronzaton. Imbs (1998) argues that 22

23 all stated problems could not be addressed just through nstrumental varables because the authentc problem s not of endogenety but of omtted varables. He beleves that unobservable components of tme-nvarable characterstcs for each country could be solved through panel estmatons wth fxed effects. In comparson between OLS and panel technque wth fxed effects, Shn and Wang (2004) conclude that the latter can elmnates the unobserved country-specfc components and t s more approprate to analyze the effect of pattern of trade on busness cycle synchronzaton. In bref, snce there s no consensus on the correct methodology to apply, all estmaton methods wll be mplemented n ths study for comparson and robustness chec as well. In addton, another econometrc method that has been not seen n emprcal lterature s also conducted n the study,.e. panel-based IV regresson. For all IV regressons, followng Franel and Rose (1998), we use some gravty varables as nstruments for trade ntensty,.e. ) log of dstance, ) dummy varable for common language and ) dummy varable for common border. All these varables are extracted from CEPII Database. 4. Estmaton Results 4.1. Statstcal Descrpton of Varables Before mplementng the regresson analyss, we present some statstcal descrpton and correlaton analyss of varables used n the models (see Tables 2, 3, 4 and 5). Ths helps provde rough ntal loos at varables and possble lns between explanatory varables wth busness cycle synchronzaton. Statstcal Summary The mean value of blateral correlaton effcent between cycles n ASEAN-5 durng the whole sample perod ( ) s 0.63 but qute hgh recently (0.88 and 0.89 for the latest sub-perods). 23

24 Table 2. Statstcal Summary of Varables Varables Mean Mn Max Std. Dev. Busness Cycle Synchronzaton Corr(y t,y jt ) Blateral Trade Intensty BTI BTI BTI BTI Intra-Industry Trade Intensty IIT Intra-Industry Trade Index Intra Intra Intra Intra Inter-Industry Trade Index Inter Inter Inter Inter Smlarty of Industry Structure SIS Smlarty of Fscal Polcy FPS Table 3. shows that blateral trade ntensty n ASEAN-5 s lely to rse over tme, especally for measures of BTI 1 and BTI 4. Ths mples an ncreasng trade ntegraton among countres. Contrbutng to hgher degree of trade ntensty, ntra-ndustry trade s observed to go up rapdly n the regon. Ths s reflected by a massve ncrease n ntra-ndustry ntensty ndex (IIT) over tme: from and n the frst two sub-perods to and n the last two sub-perods. As result, the ntrandustry trade component (Intra) s seen to ncrease, especally n recent years. Meanwhle, although ncreasng early n the whole perod, nter-ndustry trade component (Inter) s not lely to change much lately. Therefore t can be sad that trade relatonshp among ASEAN-5 has been much tghtened recently but the most mportant contrbuton s a boom n the ntra-ndustry trade. 24

25 Table 3. Statstcal Summary Over Tme Varables Corr(y t,y jt ) BTI BTI BTI BTI IIT Intra Intra Intra Intra Inter Inter Inter Inter SIS FPS The mean value of SIS varable the smlarty of ndustral structure s qute hgh: 0.70 for the whole perod ( ) and goes up across sub-perods, except the last one. Ths can be partly explaned by the fact that countres n the regon have qute smlar factor endowments and development level. Meanwhle, the fscal polcy smlarty seems to be loose. The varable FPS has the mean value of just 0.38 and s lely to have an unclear changng pattern over tme. Table 4 presents man varables (ntra-ndustry trade ntensty, blateral trade ntensty, smlarty of ndustral structure and smlarty of fscal polcy) by country over tme. For smplfcaton, only the blateral trade ntensty usng only export data (BTI 1 ) s chosen for computaton. The fndngs are consstent for other measures for ths varable. Ths wll be avalable upon request. Table 4 shows that all ASEAN-5 countres share the same ncreasng trend of ntrandustry trade ntensty over tme, especally n late sub-perods. On average, Thaland, Malaysa and Sngapore have the hghest level of ntra-ndustry trade tes whle Indonesa lacs a strong ntra-ndustry trade ln wth the remanng members. That mght be because Indonesa s a more closed economy and less ntegrated wth the rest of the regon. Ths poston s also confrmed n Cortnhas (2005) s study, 25