Macroeconomic Interdependence in East Asia

Transcription

1 MPRA Munch Personal RePEc Archve Macroeconomc Interdependence n East Asa Jun Nagayasu 1. Aprl 2010 Onlne at MPRA Paper No , posted 17. December :43 UTC

2 Macroeconomc Interdependence n East Asa Aprl 2010 Jun Nagayasu 1 Unversy of Tsukuba Abstract: Ths paper analyzes macroeconomc nterdependence among 10 Asan economes. In ths connecton, we decompose ther macroeconomc actves (real GDP) nto common and country-specfc components usng the Ba-Ng method (2004). Our results suggest frst that both components are nonstatonary and have permanent effects on ther overall economy. Second, we fnd the relatve mportance of common factors n all countres n terms of ther contrbuton to varatons n real GDP. But evdence s also obtaned of country-specfc effects becomng ncreasngly mportant n countres lke Chna n recent years. Therefore, f, for example, Chna s expected to grow at a fast pace n future, our fndngs mply that creaton of a regonal monetary unon of these 10 countres needs to be held back untl the Chnese economy has become more domnant n the regon. JEL classfcaton: F4 Keywords: Asan economc ntegraton, factor models, common and factors Fle name: Nagayasu_JWE.doc 1 Correspondng author; Unversy of Tsukuba, Graduate School of Systems and Informaton Engneerng, Tennoda, Tsukuba, Ibarak JAPAN; Tel/Fax ; Emal Nagayasu.jun.gn@u.tsukuba.ac.jp. Ths research was funded by a Grant-n-Ad for Scentfc Research (B) No An earler verson was presented at the 3rd All Chna Economcs Internatonal Conference (HK 2009) and the 3rd Internatonal Trade Workshop at the Research Center for Internatonal Economcs (Bejng 2009). I would thank for comments Juyoung Cheong, S. R. Gangadharan, Hyun-Hoon Lee, Ku-Wa L, Tatsuyosh Myakosh and other conference partcpants. However, all remanng errors are mne. 1

3 1. Introducton Ths paper focuses on regonal economc nterdependence. Hgh nterdependence s regarded as ndcatng that economes are hghly ntegrated n the regon, and ths suaton can be created by hstorcal economc and polcal efforts. Economc and polcal cooperaton has a long hstory n Asa. The Assocaton of South-East Asan Natons (ASEAN), now consstng of 10 countres, 2 was establshed n 1967 to help acheve regonal secury, and soco-cultural and economc ntegraton. Furthermore, the ASEAN Free Trade Area (FTA) agreement, whch amed to elmnate tarff and non-tarff barrers n the regon, was sgned n 1992 by sx members (Brune, Indonesa, Malaysa, Sngapore, Thaland, and the Phlppnes) and later by the others. Such regonal efforts seem to have pad off. Some Asan countres (e.g., Hong Kong (HK)) had acheved outstandng economc results for several decades and had been regarded as a world-class economc success story. However, hstory has shown that such success does not last forever. The 1997 Asan crss whch erupted n Thaland came as a surprse to many economsts and polcy-makers the world over. Its adverse effects splled over to neghborng countres, and consequently many countres experenced a sharp economc downturn. Mgatng ths contagous effect n Asan emergng markets called for further regonal cooperaton. As a result, a more comprehensve group of Asan countres, the ASEAN Plus Three Countres (Chna, Japan, and Korea), was formed n 1997 to dscuss regonal economc and fnancal stably ssues. Ths dalog led to the establshment of the Chang Ma Inatve, a short-term cred arrangement among these countres to remedy pressure from lack of foregn reserves. 3 In order to examne economc nterdependence, we shall decompose macroeconomc actves nto common and country-specfc factors, then nvestgate ther mportance, and analyze the transmsson channel whch wll create nternatonal economc nterdependence 2 The ASEAN members are Brune, Camboda, Indonesa, Laos, Malaysa, Myanmar, Sngapore, Thaland, the Phlppnes, and Vetnam. 3 See Bayoum, Echengree and Mauro (2000) and Chey (2009) for dscusson of a possble Asan monetary unon. 2

4 n the regon. In order to do ths we use the factor model proposed by Ba and Ng (2004) whch does not requre a pror assumptons about the statonary of common and country-specfc components. Ths s probably the frst tme for ths approach to be used n ths research feld, and n ths respect our research departs sgnfcantly from prevous studes whch focused largely on busness cycles (.e., economc growth) that are assumed to be statonary. Furthermore, when common factors are found to be nonstatonary, we evaluate the long-run mplcatons of economc trends n a contegrated system. Our study also dffers from most prevous studes n s use of long(er) hstorcal data ncludng post-asan crss observatons whch enable us to conduct an analyss of common factor movements n recent years. In short, we fnd evdence of more than one common factor and of both country-specfc and common components beng nonstatonary, whch, based on prevous studes, mples that both components have made a permanent contrbuton to the overall economc actves n the past. Furthermore, whle we confrm the szable contrbuton of common factors n explanng GDP varatons, the country-specfc effect seems ncreasngly mportant n some countres lke Chna snce the Asan crss. Fnally, unlke some prevous studes (see the next secton) focusng on statonary busness cycles, ths paper reports a strong relatonshp between common factors and nternatonal trade, confrmng trade as the transmsson channel between countres. 2. Lerature Revew Due to economc and polcal mplcatons, a lot of research has been attempted n order to nvestgate economc ntegraton, n partcular, usng data on busness cycles across countres. For example, Selover (2004) studed nterdependence between Korea and Japan, and found that Japanese busness cycles have a moderate effect on Korean ones. Zhang, Sato and McAleer (2004) studed a group of 10 Asan countres plus the US, compared estmates from the European Economc Communy (EEC), and concluded that underlyng structural shocks are less symmetrcal n Asa. Germany, a leadng country n Europe, shows a smlar and sgnfcant correlaton pattern of demand shocks wh other core 3

5 European countres. In contrast, Japan, a large economy n Asa, does not exhb sgnfcant correlaton wh other Asan countres. Smlarly, Moneta and Ruffer (2009) examned the output growth of 10 Asan countres from 1993 to 2005 usng a dynamc common factor model, and reported a sgnfcant common factor shared by these countres except Chna and Japan. Ther result for Japan s consstent wh Rand and Tarp (2002) who fnd that the nature and characterstcs of busness cycles n developng countres dffer from those n developed countres. However, unlke the abovementoned studes whch found that Chna (manland) and Japan are not synchronzed wh the rest of Asa, Sato and Zhang (2006) documented that Japan s one of the most ntegrated countres n the regon usng the contegraton method. Out of 55 possble pars of Asan countres, 10 pars are found to be contegrated wh a posve contegratng vector. Three pars out of 10 are related wh Japanese GDP (n levels), and nterestngly Chna s strongly correlated only wh HK. Ths somewhat dfferent result may be attrbutable to the dfferent focus of ther research: whether data are n level or dfference (.e., busness cycles). 4 Research was also conducted to try to dentfy the transmsson channel of busness cycles, and has frequently focused on nternatonal trade. 5 For example, ths confluence of busness cycles seems to be drven by a strong trade (especally export) channel (Selover 1999, Moneta and Ruffer 2009), rather than by consumpton or nvestment (Moneta and Ruffer 2009). Webber (2009) suggested that exports and nvestment are sources of common fluctuaton n Asan busness cycles. Furthermore, usng data for over 100 countres on 4 There are more comprehensve studes n terms of country coverage. Usng annual data from 106 countres from 1960 to 2005, Kose, Otrok, and Prasad (2008) reported evdence of convergence n busness cycles whn ndustral countres and whn emergng markets, but not between ndustral and emergng markets. Furthermore, they found the relatve mportance of country-specfc factors n the post-1985 perod by means of the varance decomposon method. In contrast, by decomposng busness cycles nto common and country-specfc components usng a Bayesan dynamc latent factor model, Kose, Otrok, and Wheman (2008) confrmed the ncreasng mportance of common factors n explanng varatons n output, consumpton and nvestment n the more recent perod (1986:3-2003:4) compared wh the Bretton Woods perod (1960:1-1972:2). 5 There are studes examnng fnancal market ntegraton n Asa. For example, Park and Shn (2009) documented weak evdence of fnancal ntegraton n East Asa. 4

6 nternatonal trade, ndustral structures, factor endowments, and currency unon, among others, Baxter and Kouparsas (2005) confrmed that nternatonal blateral trade s the most mportant channel. However, some researchers argue aganst nternatonal trade as a transmsson channel. For example, Crosby (2003) faled to fnd evdence to lnk between busness cycles and a trade ntensy varable n East Asa. Furthermore, Imbs (2004) underscored specalzaton patterns whch drectly reflect dfferences n GDP per capa usng data from 24 (relatvely prosperous) countres. In addon to these studes based on macroeconomc data, there s lots of research on economc ntegraton usng the mcro (frm)- level data (e.g., Ando 2006). They tend to show the hgh level of economc ntegraton n East Asa from both the mport and export sdes. However, partly due to data avalably, the research has been based almost exclusvely on the tradable goods sector lke the manufacturng ndustry. As a consequence, the non-tradable goods sector has not been covered n prevous studes. Snce our research utlzes macroeconomc data, the non-tradable goods sector s also taken nto account. 3. Statstcal Methods In order to extract country-specfc and common elements from the GDP, we use the statstcal method (Ba and Ng 2004) whch s known as the panel un root test based on a factor model. Examnaton of the tme-seres propertes of data such as statonary has pqued the nterest of many researchers over recent decades. Inally, statstcal tests (.e., the un root test) were proposed n the unvarate context and then were extended to analyze the statonary n panel data (Ln and Levn 1992). They were developed wh an assumpton of no cross-sectonal correlaton n the data. But ths s unlkely to hold n actual economc and fnancal data, and a volaton of ths assumpton bases test statstcs (O'Connel 1998). Thus Ba and Ng (2004) proposed a procedure for estmatng cross-sectonal correlaton (.e., communal elements ( F t n equaton (1) n the panel data) usng a factor model. Ths paper wll utlze these elements as a proxy for common movements across countres. Below we wll explan brefly the concept of ther statstcal approach. 5

7 For the data X ( = 1,, N and t = 1,, T ), the factor model wh ndvdual effects ( c ) can be expressed as: X = c + λ F + e t (1) where F t and e are common and country-specfc elements respectvely, and ndvduals (countres) and tme are denoted as and t. Snce these elements are unobservable, an approprate number of common factors ( r ) need to be determned by nformaton crera (e.g., Ba and Ng 2002). Gven ths nformaton, both elements are estmated by a factor model, and we can carry on testng ther statonary. However, when e s nonstatonary, the estmates of λ and they proposed a dfferencng equaton (1): x = λ f + z (2) t e are no longer consstent, and therefore where x = X, ft Ft =, and z = e. Equaton (2) suggests that f t s common to all ndvduals, but λ makes a unque level of common factors λ f t for each country. Thus, one can nterpret λ as a parameter for capturng the nfluence of f t over the countres to a dfferent degree. They are a reasonable proxy for common factors because for example some countres are more affected by ol shocks than others, and the extent to whch the country s affected by ths shock can be measured by as λ λ. We call λ f t as well F t common factors n the subsequent study, and thus our concept of the common factor may be slghtly dfferent from other studes that assume constant common factors across countres. Fnally, where t = 2,, T. e and F t can be recovered by e t t = s=2 zs and Ft = s=2 fs Whle our focus s more on common factors, we are also nterested n country-specfc components. Testng ndvdually the statonary of the country-specfc component for country s dentcal to the standard Augmented Dckey-Fuller (ADF) test based on equaton (2). 6

8 ê = d 0 ê 1 + d 1 ê d ê error (3) p p + where estmates are denoted by. Based on ths test for ndvdual countres, we can calculate the statstc for evaluatng the statonary of a group of country-specfc components by poolng p -values ( p e ( )) obtaned from the ndvdual ADF test. P N = ln( pe( )) 2N N( 0, ) (4) 4N 2 1 e = 1 Ths statstc s shown to be asymptotcally normally dstrbuted, and s large posve value becomes evdence aganst the null of no contegraton. Wh respect to common factors, testng ther statonary s dentcal to the standard ADF n the presence of a sngle common factor. If we assume that changes n common factors contan the constant, the test s based on the followng equaton. Fˆ = c0 + δ0fˆ t 1 + δ1 Fˆ t δ p Fˆ t p error (5) t + In ths case, the statstc s referred to as ADF f, and the crcal value equals at the fve percent sgnfcance level. In the presence of multple common factors, one can use the multvarate contegraton method to check f there s a long-run relatonshp between the common factors. In ths paper, we use the Johansen test whch s probably the most popular multvarate contegraton method. 4. Data We consder 10 Asan economes: Chna (manland), Hong Kong (HK), Indonesa, Japan, Malaysa, Sngapore, South Korea, Tawan, Thaland, and the Phlppnes, fve of whch are ASEAN members. In addon, the US s ncluded n our data set snce s an mportant tradng partner of all these countres. Long hstorcal data on the GDP for most Asan countres are not readly avalable, and therefore, we obtaned quarterly real GDP data from Tlak Abeysnghe's homepage ( n order to evaluate common factor movements. We utlze all countres lsted there and create a balanced panel data set spannng from 1975Q1 to 2007Q1 (base year = 1995). 7

9 These data are plotted n Fgures 1 and 2, and ther basc statstcs are summarzed n Table 1. Real GDP n Fgure 1 shows that there s a sgnfcant economc slowdown n 1997 at the tme of the Asan crss but there are sgns of a prompt economc recovery. A smlar trend n real GDP growth can be observed n Fgure 2. Table 1 summarzes the basc statstcs of real GDP (both n levels and dfferences). Accordng to ths table, the Chnese economy (GDP n levels) s very volatle as her standard devaton s far hgher than others, and by contrast Japan, the Phlppnes and US experenced very low volatly. In addon, the correlaton matrx s presented n Table 2. It shows that most pars are posvely correlated wh each other regardless of the data beng the level or growth of GDP. However, unlke other countres, the Chnese GDP growth s negatvely correlated wh other countres ncludng Japan, Malaysa, South Korea, Tawan, Thaland and the Phlppnes. Ths ndcates that, as prevous studes have suggested, the Chnese economy may be less ntegrated wh the rest of Asa. We also note that all Asan countres except the Phlppnes are posvely correlated (but nsgnfcantly) wh the US GDP growth. Wh respect to trade data, three sources are used. Quarterly total mport and export data are obtaned from the IMF's Internatonal Fnance Statstcs (IFS). Ther real values are calculated usng the consumer prce ndex (CPI) from the IFS. In addon, quarterly blateral export and mport data are obtaned from the Drecton of Trade (DOT) data set also from the IMF. But coverage s lmed to Chna, HK, Indonesa, Japan, Malaysa, Sngapore, South Korea, Thaland, and the Phlppnes, and thus n order to supplement them total mport and export data for Tawan are downloaded from the homepage of the Central Bank of the Republc of Chna (Tawan). 5. Emprcal Results 5.1. PANIC estmates The results of decomposon of real GDP to common and country-specfc factors are summarzed n Table 3. Dfferent groupngs of countres are consdered n order to check the robustness of our fndngs. Our benchmark model conssts of 10 Asan countres ( N = 10 ). The group of 9 ( N = 9 ) drops Chna from our benchmark, and that of 11 8

10 ( N = 11) adds the US to the group of 10. Frst, we calculate the number of common factors ( r ) usng nformaton crera, IC1, IC2, and IC3 (Ba and Ng 2002), and fnd that there s evdence of one and three common factors dependng on the nformaton crera. A sngle common factor s supported by IC3, and three factors by IC1 and IC2. Ths result s generally robust to the composon of the panel of countres ( N = 9, 10 or 11). As there s at least one common factor, appears essental for these Asan countres to consder the economc condons of ther neghborng countres when formng economc stablzaton polces. Our estmates of common ( λ F t ) and country-specfc ( e ) factors for r = 3 and N = 10 are plotted n Fgures 3 and 4 respectvely. Fgure 3 shows a sharp fall n the common factors around 1997, whch suggests that the adverse effects of the Asan crss were shared by these countres. Fgure 4 shows that the country-specfc effect s becomng ncreasngly mportant n Chna. Snce these values are cumulatve values of λ f t and respectvely (see Secton 3), there are only small dfferences n ther nal values among countres. We also check the statonary of these factors ( λ F t and z e ), and the results are reported n Table 3. Our statstc ( P e ) n equaton (4) shows that a group of country-specfc factors s nonstatonary, and smlarly each common factor s found to be nonstatonary. Snce multple common factors are found n our data, s of nterest to examne the statonary of the group of common factors. We test ths by means of the Johansen multvarate contegraton test usng dfferent composons of common factors (.e., r = 2, 3) and report evdence of non-contegraton n Table 4. The nonstatonary of both factors suggests that country-specfc and common shocks are both permanent, and that both elements are mportant n determnng ther long-run economc performance. Therefore, ths suggests that the nonstatonary of real GDP reported n prevous studes (e.g., Sato and Zhang 2006) s attrbutable to the nonstatonary of both factors The Relatve Importance of Common and Country-Specfc Factors 6 It follows that country-specfc and common factors are not contegrated. 9

11 The relatve mportance of common factors s examned by calculatng three ratos: 1) the rato of the standard devaton of common factors to that of GDP, 2) the rato of the standard devaton of country-specfc factors to that of GDP, and 3) the rato of country-specfc factors to that of common factors. Where common factors are relatvely mportant, the frst rato should approach one, and the second and the thrd ratos should be close to zero. On the other hand, f country-specfc factors are domnant, the second rato should approach one. Whle s certanly a smplstc method, helps us understand ther relatve mportance. The subsequent analyss s based on three common factors ( r = 3) whch two out of three nformaton crera suggest. Table 5 summarzes the results of these ratos wh dfferent assumptons about the composon of countres ( N ) n the panel. 7 Frst, there s no doubt that common factors are mportant n all countres, and they seem to domnate GDP varatons partcularly n Sngapore and South Korea. Ther frst rato s close to one, and the second and thrd ratos measurng the contrbuton of country-specfc effects are nomnal. The frst rato of HK and the Phlppnes s also close to one, but ther second and thrd ratos are relatvely larger than that of Sngapore and South Korea. Indeed, the mportance of the common factor s confrmed by the varance rato test. Although there reman sgnfcant dfferences between e and X, the varance rato test for ( λ F, X ) suggests that, except Chna whose frst rato s around 0.8, the varaton of the common factor s statstcally and consstently dentcal to that of the total GDP. Ths result seems to be generally robust to the composon of countres under nvestgaton. Gven that these economes experenced a transon phase en route to ndustralzaton, we also check f the relatve mportance of the factors has changed over tme. Table 6 shows the p -values of the varance rato tests for these ratos n two sub-sample perods when r = 3. 8 The breakng pont of 1997Q2 s consstent wh the economc dsaster n Thaland, 7 Our dfferent assumptons are based on there beng one, two or three common factors and a group of nne, ten or eleven countres, but such assumptons seem to barely alter the fnal results. For ths reason, such results are not reported here. 8 The results wh r = 1 and 2 are not reported here due to space constrants. But the results wh a 10

12 the frst country h by the Asan crss. Generally, we can observe a smlar pattern n these ratos to those from the full sample. However, there s evdence that country-specfc effects become ncreasngly sgnfcant n the post-1997 perod, partcularly, n Chna and the Phlppnes. The null of the varance test for ( λ F, e ) cannot be rejected for these two countres n the pre-1997 perod, but can be rejected n the post-1997 perod. The rapd economc growth n Chna n recent years and the lesser effects of the 1997 crss on the Phlppnes (Fgure 2) may be attrbutable to ths result The Transmsson Channel Whle many other transmsson channels can be consdered (see Secton 2), ths sub-secton focuses on nternatonal trade as a transmsson channel of economc actves due to data avalably, and looks at whether there s a posve and contegrated relatonshp between common factors ( λ F t ) and nternatonal trade n the panel data context. The presence of contegraton ensures that there s a lnear combnaton between them and becomes evdence of a long-run relatonshp. We would expect that common factors and trade related data are posvely correlated and also contegrated snce many Asan countres have adopted an open market polcy and nternatonal trade has been regarded as an engne of economc development. But as dscussed, some prevous studes (e.g., Crosby 2003, Imbs 2004, Shn and Wang 2004) queston the role of (smple) nternatonal trade as a transmsson channel of busness cycles, and argue that s other characterstcs of a country such as the ntra-ndustry trade and ndustry structures whch create nterdependence n busness cycles. Here we use two types of trade data. One s the real value of total mport and export data, and the other s trade concentraton measures whch are also created separately for mports and exports. 9 dfferent sze of r wll not alter our general concluson. 9 A smlar defnon of concentraton ratos was used prevously (e.g., Frankel and Rose 1998; Shn and Wang 2004), and these ratos are calculated for each country and tme perod. 11

13 Im p Im p jt for mports and Exp Exp jt for exports where Im p and Exp refer to mports and exports respectvely. The subscrpts ( and j ) represent home country ( ) and the rest of Asa ( j ), and t s tme. Thus Exp jt / Exp shows the exports of country to the rest of Asa dvded by the total exports of country, and Im p / Im p ndcates the proporton of mports to country from Asa to the jt total mports to that country. Here, we use the benchmark model, and thus Asa s defned as the 10 Asan countres used n ths study. Snce hgh ratos ndcate hgh concentratons of regonal trade, one mght expect that a hgh concentraton rato would be closely and posvely assocated wh common factors. Table 7 summarzes the trade data and shows that regonal trade whn ASEAN s hgh; the trade concentraton measure of most ASEAN countres s around 50 percent. The non-aseans lke Japan and Korea exhbed a slghtly lower level of regonal trade, but nterestngly, manland Chna shows a hgh concentraton of nternatonal trade wh other Asan countres. 10 Table 8 reports the estmated relatonshp, based on r = 3 and N = 10, between the common factor and trade-related data. 11 We use several panel data estmaton methods (OLS, Adjusted (Adj.) OLS, and Fully Modfed (FM) OLS). 12 Generally, when total trade data s employed, a posve and sgnfcant relatonshp s obtaned for both mports and exports, and there s contegraton between the trade value and common factors. However, when the trade concentraton s consdered, evdence to support ther relatonshp wh common factors (n levels) becomes very weak; there s no evdence of contegraton between them usng the panel contegraton method (Kao 1999). There s a possbly that structural breaks may destroy the relatonshp, but our result s also confrmed by the 10 Ths table also shows that the US s an mportant trade partner for all countres, especally for Japan, Korea and the Phlppnes. 11 Prevous studes (e.g., Wu, Chen and Lee 2001) often showed that export and mport data are nonstatonary. 12 See Kao and Chang (2000). 12

14 Westerlund test (2006) whch takes account of multple and unknown structural breaks n the panel data (see Appendx). 13 These mport and export elastces seem stable over tme. Fgure 5 shows that estmates of mports and exports usng the Asan related trade data by extendng one observaton by one observaton from 1997Q2 to 2007Q1. The sensvy of trades sharply dropped rght after the erupton of the Asan crss wh the excepton of the estmate of mports n 1997, and has become very stable snce then. In addon, we can observe that the estmate of exports s always hgher than that of mports, confrmng the relatve mportance of exports n explanng the common factors. Therefore, we conclude that trade s one mportant transmsson channel for nternatonal economc nterdependence, and the result mples that our common factors are also nfluenced by economc developments elsewhere such as the US and Europe. Furthermore, Chna whch exhbs hgh trade concentraton wh other Asan countres (Tables 7) shows a low correlaton wh common factors (Tables 5 and 6). Ths mples the sgnfcance of her domestc market, and therefore although Chna trades a lot wh other Asan countres, one may conclude that nternatonal trade s less mportant for her compared wh other Asan economes. In general, our results strongly support the role of nternatonal trade, and n partcular, both mports and exports are found to be a drvng force of common factor movements. Ths pont was controversal n prevous studes usng busness cycle data, but a statstcally more sensble approach seems to yeld our rather clear-cut result. Use of common factors whch presumably contan more nternatonal elements than whole economc actves (.e., busness cycles) and consderaton of ther tme-seres propertes appear to contrbute to the stronger evdence n favor of ths relatonshp. 6. Summary and Dscusson 13 In order to check ths weak relatonshp between the concentraton rato and common factors, we also employ the growth (rather than the level) of the common factors. The results are not shown n the paper due to lmed space, but the parameter sgn remans generally unchanged although the lnear relatonshp wh the growth of the common factors s now found to be statonary. 13

15 We studed economc dependence n macroeconomc actvy among East Asan countres. Monorng ther level of regonal ntegraton s mportant not only when consderng further economc and fnancal ntegraton but also when formng economc polcy to stablze ther own economes. When a country s largely dependent on other members, a shock n one member country wll drectly and possbly quckly nfluence other countres. Our results are as follows. Usng real GDP data and decomposng macroeconomc actves nto the common and dosyncratc factors by means of the Ba-Ng method (2004), we fnd more than one common factor among Asan countres. Furthermore, common factors domnate varatons n GDP n each country, confrmng ther open economc polcy; n other words, nternatonal trade s a drvng force of the common factors. In addon, whle ther sze may be nconsequental, both common and country-specfc factors have a permanent effect on macroeconomc actves (.e, they are nonstatonary), and country-specfc factors are ncreasngly sgnfcant n recent years n partcular n Chna and the Phlppnes. Ths may underlne the sgnfcant presence of non-tradable sectors. Unlke the condon for nternatonal trade, monetary unon requres homogeneous economc envronments. Therefore, f for example Chna s expected to grow at a fast pace n future, our fndngs mply that creaton of a regonal monetary unon of these 10 countres needs to be delayed untl the Chnese economy has become more domnant n the regon. Fnally, whle our analyss s statstcally sold, there are many ssues that one could nvestgate n the future. For example, we focused on nternatonal trade as a transmsson channel of common factors. However, there are many other channels through whch stocks are transmted across countres (See Secton 2). Ths can be carred out when more data are dssemnated for these countres. We beleve that the understandng and dentfcaton of the exact nature of transmsson mechansms wll help propose a more concrete approach for further economc ntegraton. 14

16 References Ando, M., Fragmentaton and vertcal ntra-ndustry trade n East Asa. North Amercan Journal of Economcs and Fnance 17, Ba, J., Ng, S., Determnng the number of factors n approxmate factor models. Econometrca 70(1), Ba, J., Ng, S., A Panc attack on un roots and contegraton. Econometrca 72(4), Baxter, M., Kouparsas, M. A., Determnants of busness cycle movements: a robust analyss. Journal of Monetary Economcs 52(1), Bayoum, T., Echengreen, B., Mauro, P., On regonal monetary arrangements for ASEAN. Journal of the Japanese and Internatonal Economes 14, Chey, H-K., A polcal economc crque on the theory of optmum currency areas, and mplcatons for East Asa, World Economy 32, Crosby, M., Busness cycle correlaton n Asa-Pacfc. Economc Letters 80, Frankel, J. A., Rose, A. K., The endogeney of the optmum currency area crera. Economc Journal 108, Gales, A., Sgherr, S., Regonal fnancal spllovers across Europe: a global VAR analyss. IMF Workng Paper WP/09/23. Gregory, A. W., Hansen, B. E., Resdual-based tests for contegraton n models wh regme shfts. Journal of Econometrcs 70, Imbs, J., Trade, fnance, specalzaton and synchronzaton. Revew of Economcs and Statstcs 86(3), Kao, C., Spurous regresson and resdual-based tests for contegraton n panel data. Journal of Econometrcs 90, Kao, C., Chang, M. H., On the estmaton and nference of a contegrated regresson n panel data. Advances n Econometrcs 15, Kose, M. A., Otrok, C., Wheman, C. H., Internatonal busness cycles: world, regon, and country-specfc factors. Amercan Economc Revew 93(4), Kose, M. A., Otrok, C., Prasad, E. S., Global busness cycles: convergence or decouplng? NBER Workng Paper No Kose, M. A., Otrok, C., Wheman, C. H., Understandng the evoluton of world busness cycles. Journal of Internatonal Economcs 75, Levn, A, Ln, C-F., Un root tests n panel data: Asymptotc and fne-sample propertes. Dscusson Paper 92-23, Unversy of Calforna, San Dego. Moneta, Fabo and Rasmus Ruffer, 2009, Busness cycle synchronzaton n East Asa, Journal of Asan Economcs 20, Park, Y. C., Shn, K., Economc ntegraton and changes n the busness cycle n East Asa: s the regon decouplng from the rest of the world? Asan Economc Papers 8(1), Rand, J., Tarp, F., Busness cycles n developng countres: are they dfferent? World Development 30, Sato, K., Zhang, Z., Real output co-movements n East Asa: any evdence for a Monetary Unon? World Economy,

17 Selover, D. D., Internatonal nterdependence and busness cycle transmsson n ASEAN. Journal of the Japanese and Internatonal Economes 13, Selover, D. D., Internatonal co-movements and busness cycle transmsson between Korea and Japan. Journal of the Japanese and Internatonal Economes 18, Shn, K., Wang, Y., Trade ntegraton and busness cycle synchronzaton n East Asa. Asan Economc Papers 2(3), Stockman, A., Sectoral and natonal aggregate dsturbances to ndustral output n seven European countres. Journal of Monetary Economcs 21, Zhang, Z., Sato, K., McAleer, M., Is a monetary unon feasble for Asa? Appled Economcs 36, Weber, E., Common and uncommon sources of growth n Asa Pacfc. Journal of the Japanese and Internatonal Economes 22, Westerlund, J., Testng for panel contegraton wh multple structural breaks. Oxford Bulletn of Economcs and Statstcs 68(1), Wu, L-L., Chen, S-L., Lee, H. Y Are current account defcs sustanable? Evdence from panel contegraton. Economcs Letters 72,

18 Appendx The Westerlund LM test (2006) s for detectng contegraton wh multple and unknown structural breaks n the panel data based on the LM method. The notable features of ths method nclude s ably to detect unknown multple structural breaks whch are consdered under both the null and alternatve hypotheses, and multple structural breaks are allowed n the determnstc component. More precsely, hs model evaluates the null hypothess of contegraton: H 0 : ϕ = 0 for all = 1,, N aganst the alternatve of no contegraton: H1 : ϕ 0 for = 1,, N1 and ϕ = 0 for = N1 + 1,, N, usng the followng specfcaton. Ln( Common ) = zγ + Ln( Trade ) β + e e = r + u r = r + ϕ u 1 where subscrpt t ( t = 1,, T ) represents tme and ( = 1,, N ) prefectures. z s the vector of the determnstc term ( z = [ 1] ) whch vares between regmes, and the resdual, u, s assumed to be statonary. Greek letters are parameters to be estmated. Ths test essentally examnes the tme-seres propertes of the resdual, e. When ϕ = 0 and r = 0 0 whch s a reasonable assumpton n the presence of the fxed effects, r = 1 0 and thus r = 0. Ths suggests that e = u and that there s contegraton n the panel data snce s assumed to be statonary. Ths LM test statstc requres frst the dentfcaton of the number of regmes ( j = 1,..., M + 1 and so M s the maxmum number of structural breaks, where = 1,..., N ) and of break ponts ( T = ( 1,..., ) T TM ). These notatons mean that T = 1 0 and TM +1 = T, and break ponts for each equaton are determned by mnmzng the sum of squared resduals. Ths s estmated for each regme and snce all combnatons of j and T are consdered, s called a global mnmzer. Gven the value of for each j, the optmal number of structural breaks s determned usng the Schwarz Bayesan nformaton creron. Gven the nformaton about the sze and locaton of the breaks, we can construct the LM test statstcs. 17

19 where 2 ω 11 s the long-run varance of ( e, v ) where v = x x 1, and = t S k = T + e j 1 1 k where e k s the effcent estmate of e whch can be estmated by the Fully-Modfed OLS. Wh some adjustment terms, Westerlund shows that ths statstc follows the standard normal dstrbuton, and s large value suggests rejecton of the null. 18

20 Tables Table 1. Basc Statstcs of Real GDP (log) Levels Mean Standard Max Mn Devaton Chna HK Indonesa Japan Malaysa Sngapore S. Korea Tawan Thaland Phlppnes USA Dfferences Chna HK Indonesa Japan Malaysa Sngapore S. Korea Tawan Thaland Phlppnes USA Note: Full sample. 19

21 Table 2. Correlaton Matrx (log GDP, Full Sample) Levels Chna HK Indonesa Japan Malaysa Sngapore S Korea Tawan Thaland Phlppnes Chna HK Indonesa Japan Malaysa Sngapore S. Korea Tawan Thaland Phlppnes USA Dfferences Chna HK Indonesa Japan Malaysa Sngapore S. Korea Tawan Thaland Phlppnes USA

22 Table 3. PANIC Test Results Full Sample Country-specfc factor, Common factor, Informaton Crera, P ADF IC1 IC2 IC3 e N = , , N = , , N = , , Note: Full sample. The PANIC test and nformaton crera are based on Ba and Ng (2004) and Ba and Ng (2002) respectvely. The statstc ( P e ) for evaluatng the statonary of country-specfc factors s normally dstrbuted and thus s 5% crcal value s When there s only one common factor, the factor un root test ( ADF ) has a 5% crcal value of (the constant only). A maxmum of 4 common factors are consdered when decdng the true number of common factors n the nformaton crera. N = 10 refers to all 10 Asan countres, and N = 9 to 9 when Chna s not ncluded. N = 11 s the panel of 10 Asan countres and the US. f f Table 4. Contegraton among Common Factors by Johansen Test N = 10 Trace statstcs ( p -value) N = 9 Trace statstcs ( p -value) N = 11 Trace statstcs ( p -value) Factors 1, 2 r = (0.358) (0.523) (0.299) r = (0.131) (0.063) (0.138) Factors 1, 2, 3 r = (0.259) (0.505) (0.199) r = (0.447) (0.332) (0.449) r = (0.117) (0.061) (0.109) Note: Full sample. The r s the number of common factors, and N s that of countres. The lag length of four s used for ths test. 21

23 Table 5. The Relatve Importance of Common Factors (Full Sample) r = 3, N = Std (λf)/std (X ) Var rato test (λf, X) (p-value) Chna HK Indonesa Japan Malaysa Sngapore S.Korea Tawan Thaland Phlppnes US Std (e )/Std (X ) Var rato test (e, X) (p-value) Chna HK Indonesa Japan Malaysa Sngapore S.Korea Tawan Thaland Phlppnes US Std (e )/Std (λf) Var rato test (e, λf) (p-value) Chna HK Indonesa Japan Malaysa Sngapore S.Korea Tawan Thaland Phlppnes US Note: Includes the constant term. X s real GDP. The varance (Var) rato test. 22

24 Table 6. The Relatve Importance of Common Factors (Varance Rato Test(p-value) wh Dfferent Sample Perods) r = 3, N = Q2-1997Q2 1997Q3-2007Q1 (λf, X) (e, X) (e, λf) (λf, X) (e, X) (e, λf) Chna HK Indonesa Japan Malaysa Sngapore S.Korea Tawan Thaland Phlppnes r = 3, N =9 HK Indonesa Japan Malaysa Sngapore S.Korea Tawan Thaland Phlppnes r = 3, N =11 Chna HK Indonesa Japan Malaysa Sngapore S.Korea Tawan Thaland Phlppnes US Note: See table 5. 23

25 Table 7. Trade Concentraton (Average) Imports Exports Whn Asa Wh the US Whn Asa Wh the US Chna HK Indonesa Japan Malaysa Sngapore S. Korea Tawan Thaland Phlppnes Note: The data are from the IMF's Drecton of Trade, and statstcs are computed as trade to/from each partner country dvded by the total trade of the home country. The sample perod s 1981Q1 to 2007Q1. 24

26 Table 8. The Long-Run Relatonshp between Trade and Common Factors Total trade (p-value) Trade whn Asa (p-value) Trade concentraton (p-value) Import Export Import Export Import Export OLS Adj OLS FMOLS Panel contegraton test * DF ρ * DF t (0.126) Panel contegraton test wh breaks LM (0.157) (0.142) (0.304) (0.435) Note: The panel contegraton test s based on Kao (1999) whch examnes the null of no contegraton, and the lag length of four s determned by the Akake Informaton Creron. The panel contegraton test wh structural breaks s based on Westerlund (2006) whch examnes the null of contegraton wh breaks, and breaks are consdered n the constant and tme trend. The maxmum number of breaks s three. Fgures n parentheses are p -values. Due to data avalably, the analyss usng total trade data s based on 10 Asan countres, whle that for trade concentraton s based on 9 countres (.e., excludng Tawan). The sample perod s 1981Q1 to 2007Q1. 25

27 Fgures Fgure 1. Standardzed Real GDP of Asan Countres (n natural log) Fgure 2. Real GDP Growth (log dfference) 26

28 Fgure 3. Common Components (λf) Note: See the man text (Secton 3) about the defnon and dervaton of λf. Fgure 4. Country-specfc (e) Components Note: See the man text (Secton 3) about the defnon and dervaton of e. 27

29 Fgure 5. The Parameter Sze for Imports/Exports Note: The mports and exports represent the total amount of trade to and from Asan countres. 28