A factor-augmented VAR analysis of business cycle synchronization in east Asia and implications for a regional currency union

International Review of Economics and Finance 39 (2015) 449–468

Contents lists available at ScienceDirect

International Review of Economics and Finance journal homepage: www.elsevier.com/locate/iref

A factor-augmented VAR analysis of business cycle synchronization in east Asia and implications for a regional currency union Hyeon-seung Huh a, David Kim b, Won Joong Kim c,⁎, Cyn-Young Park d a

School of Economics, Yonsei University, Republic of Korea School of Economics, University of Sydney, Australia c Department of Economics, Konkuk University, Republic of Korea d Economic Research and Regional Cooperation Department, Asian Development Bank, Philippines b

a r t i c l e

i n f o

Article history: Received 23 July 2014 Received in revised form 22 July 2015 Accepted 22 July 2015 Available online 4 August 2015 JEL classification: E32 F33 F44 Keywords: Business cycle synchronization Asian currency union Factor-augmented VAR

a b s t r a c t Debate continues over whether a monetary or currency union will be a viable alternative to the current exchange arrangements in East Asia. This paper adds to the literature by assessing the level of business cycle synchronization among 10 major East Asian countries, which is considered as a key precondition for a regional currency union. Unlike previous studies, this paper employs a factor-augmented VAR model that encompasses a large set of 62 foreign and domestic variables simultaneously. Five common shocks are identified, and we examine how and to what extent these shocks affect each economy in the region. Empirical results indicate that the majority of East Asian countries respond similarly to world and regional shocks. Of particular importance is the finding that individual GDPs are well synchronized in response to the two major determinants of world and regional GDP shocks. Overall, the evidence presents a positive case for consideration of a regional currency arrangement in East Asia. Some suggestions are offered concerning steps to build a foundation towards greater monetary cooperation in East Asia. © 2015 Elsevier Inc. All rights reserved.

1. Introduction East Asia has become one of the fastest growing and most dynamic regions in the world over the last half a century. In the 1980s and 1990s, leading up to the Asian financial crisis of 1997–98, economies in the region kept stable exchange rates, especially vis-à-vis the U.S. dollar, to support exports, a key driver of growth for many of them. However, “de facto” currency pegs against the U.S. dollar, which allowed a build-up of macro-imbalances, led to the crisis. In its aftermath, East Asian countries were forced to transition towards more flexible exchange rate regimes. However, freely floating exchange rates imply excessive bilateral volatility in the region considering the extent of its trade integration; this fact prompted a number of research and policy debates over alternative exchange rate arrangements to promote stability and credibility in exchange rates. The launch of the euro in 1999 also simulated great interest in regional monetary integration or even a currency union for many economies, including those in East Asia.1

⁎ Corresponding author at: Department of Economics, Konkuk University, 120 Neungdong-ro, Gwangjin-Gu, Seoul 143-701, Republic of Korea. Tel.: +822 450 0530; fax: +822 446 3615. E-mail addresses: [email protected] (H. Huh), [email protected] (D. Kim), [email protected] (W.J. Kim), [email protected] (C.-Y. Park). 1 The interest was somewhat watered down since the recent European debt crisis.

http://dx.doi.org/10.1016/j.iref.2015.07.010 1059-0560/© 2015 Elsevier Inc. All rights reserved.

450

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

Along with these developments, East Asian countries stepped up efforts to enhance their financial stability through financial and monetary cooperation in the region. In 1999, ASEAN expanded to include the region's larger neighbors: China, Japan, and Korea (ASEAN+3). In 2000, ASEAN+3 announced the establishment of the Chiang Mai Initiative (CMI), a network of bilateral swap arrangements, in an effort to prevent another crisis. In 2010, in part driven by the desire to fend off panic from the global financial crisis of 2008–09, the CMI was expanded and transformed into a multilateral arrangement, the Chiang Mai Initiative Multilateralization (CMIM). It draws from a foreign exchange reserve pool worth US$120 billion, which was expanded to $240 billion in 2012. In 2011, ASEAN+3 also established the ASEAN+3 Macroeconomic Research Office (AMRO) in association with CMIM to strengthen regional macroeconomic surveillance. In parallel, economic integration has accelerated in the region through increasing trade and business linkages. ASEAN envisions the creation of the ASEAN Economic Community (AEC) by 2015. This paper aims to empirically assess the feasibility of a currency union in East Asia. A regional monetary or currency union presents an economic trade-off between the potential costs and benefits of joining it.2 The main expected advantages are exchange rate stability, a reduction in transaction costs, price convergence and stability, and the consolidation of regional markets. A currency union also helps enhance national economic policy discipline due to the regional economic surveillance and monitoring arrangements that go along with it. The problems associated with a currency union arise from the diversity of the economies joining it. When economies with different economic fundamentals, levels of efficiency, and levels of productivity are joined under a single currency, a single form of monetary control may not be suitable to address the disparate domestic economic conditions. A loss of national monetary control implies that the national monetary authorities can no longer undertake measures to address country-specific macroeconomic conditions, such as combating inflation or reducing unemployment.3 The member countries will also lose nominal exchange rates as an adjustment mechanism in response to country-specific external shocks such as swings in foreign demand or sudden interruptions in capital inflows. Despite the progress of regional economic integration, there is substantial heterogeneity across East Asian countries in the level of economic and financial development. The skepticism about a common currency area has been further fuelled by recent economic turmoil in the euro zone. Indeed, the feasibility of a common currency for East Asia has been a topic of continued debate. The theory of optimal currency area (OCA), introduced by Mundell (1961) and McKinnon (1963), provides a list of important criteria for a common currency in a region. According to traditional OCA theory, countries may consider adopting a common currency when their economies have (i) similar shocks and business cycles, (ii) high trade integration, (iii) internal labor mobility, and (iv) internal fiscal transfers. Among these factors, synchronization of business cycles is the most well understood and is considered a key precondition. Once a currency union is instituted, member countries will follow a common monetary policy established by a super-national central bank. If they are less synchronized in their business cycles, the cost of surrendering monetary policy autonomy is bound to be significant, as they cannot adopt independent policies suited to their domestic conditions. If the business cycles are assimilated, the cost tends to be low or negligible and can be outweighed by the benefits of forming a currency union. In this context, the present paper empirically investigates the degree and economic nature of business cycle synchronization among the 10 most economically significant countries in East Asia with the aim of uncovering ‘hard’ evidence concerning the viability of a currency union in the region. The selected countries are the ASEAN5 (Indonesia, Malaysia, Singapore, Thailand, and the Philippines) as well as China, Hong Kong SAR, Japan, Korea, and Taiwan. For each country, six key macroeconomic variables are employed to characterize business cycle movements: GDP, the real exchange rate, inflation, money growth, exports, and imports. World oil prices and world GDP are also included to reflect changes in the world market; thus, the model comprises a total of 62 variables. The paper models this large dataset and evaluates the dynamic interactions among variables using the factor-augmented VAR (FAVAR) approach developed by Forni, Hallin, Lippi, and Reichlin (2000), Bernanke, Boivin, and Eliasz (2005), and Stock and Watson (2005). FAVAR provides a parsimonious and effective framework of analysis by extracting common factors from a large set of data and utilizing the benefits of VAR models in characterizing the dynamics. This approach is also useful when the time span of the data is rather short, as in this case. The structural shocks underlying the model are classified as global, regional, and idiosyncratic shocks, as economies can respond differently to shocks depending on their origin. By construction, each country is subject to common world and regional shocks. This allows us to compare the responses of the variables to these shocks across countries, which in turn facilitates a straightforward interpretation of business cycle synchronization. Our study is closely related to those of Eickmeier (2009) and Bagliano and Morana (2009) in terms of its objectives and methodology. Using FAVAR models, they documented the co-movements of key macro-variables in the core Eurozone countries and in the U.S., U.K., Canada, Japan, and the euro area. Most of other studies have used small-scale VAR or factor models to assess the degree of business cycle synchronization as a means of determining the suitability of a currency union. While there is a large body of literature on the European business cycle, the parallel literature on the Asian business cycle has been modest. Bayoumi and Eichengreen (1994) undertook a benchmarking study of Asian business cycle synchronization. They estimated a bivariate VAR of GDP and inflation to identify supply and demand shocks using a long-run restriction approach of the type proposed by Blanchard and Quah (1989). Taking the correlation of identified shocks between each pair of countries as a measure of business cycle synchronization, they suggested which group of East Asian countries may consider forming a currency union in the region. Subsequently, a number of studies followed suit, including Bayoumi, Eichengreen, and Mauro (2000), Yuen (2001), Kim (2007), Rana (2007) and Lee and Koh (2012). However, Lee, Park, and Shin (2003) and Lee and Azali (2012) raised several issues concerning such studies. They argue that an examination of the region-wide synchronization, rather than bilateral data, is more appropriate because the net benefits of adopting a currency union need to be assessed across the entire set of economies in a region. Moreover, underlying shocks may be estimated differently in each 2 3

While currency union and monetary union are not identical as monetary union is broader in its scope, the two terms will be used interchangeably in this paper. A further cost of losing domestic monetary controls is the loss of seigniorage.

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

451

country, making it difficult to compare the results across countries with the same base. Finally, a standard correlation analysis of shocks does not account for the sources of the shocks. A third factor, such as a world income shock, can induce a high correlation between countries. To overcome these shortcomings, Lee et al. and Lee and Azali proposed to employ a dynamic factor model similar to that in Kose, Otrok, and Whiteman (2003). This model examines the degree of business cycle synchronization among a group of countries in the presence of world, regional, and idiosyncratic shocks. They concluded that regional shocks explain the movements in the outputs of East Asian countries and that the region is well prepared for a currency union based on that criterion. Using similar models, Moneta and Rüffer (2009) reported that most East Asian countries share significant common growth dynamics, while He and Liao (2012) found strong evidence of Asian business cycles being independent of U.S. and European cycles. While these factor models simultaneously capture the outputs of countries in the region, the computational complexities associated with maximum likelihood estimations can hinder one from accounting for many different types of variables and thus accommodating richer dynamics. Even when this is possible, the factors identified can often lack structural interpretations. We employ a factor-augmented VAR model that encompasses a large set of 62 foreign and domestic variables simultaneously. The paper identifies common economic shocks to the economies and examine how and to what extent these shocks affect each regional economy. Empirical results suggest that the majority of East Asian countries respond similarly to world and regional shocks. We also find that individual GDPs are well synchronized in response to the two major determinants of world and regional GDP shocks. The rest of this paper is organized as follows. Section 2 briefly presents the FAVAR model used for the application at hand. The data are described in Section 3, and Section 4 details the extraction of common factors that summarize co-movements in the model. Section 5 explains how East Asian countries respond to underlying common shocks with a view to assess the synchronization of business cycles in the region. Finally, Section 6 discusses the implications of our findings with regard to the feasibility of a currency union in East Asia and concludes the paper. 2. Econometric model and estimation 2.1. A FAVAR model It is assumed that the joint dynamics of n variables are specified by the following dynamic factor model (Bagliano & Morana, 2009; Stock & Watson, 2005),4 X t ¼ Λ F t þ DðLÞX t−1 þ νt

ð1Þ

F t ¼ ΦðLÞF t−1 þ ηt ;

ð2Þ

where Xt is an (n × 1) vector of stationary variables, Ft is an (r × 1) vector of unobserved common factors with r b n, Λ is an (n × r) matrix of the loading coefficients, D(L) is an (n × n) lag polynomial of order p, and Φ(L) is an (r × r) lag polynomial of order q. vt is an (n × 1) vector of idiosyncratic disturbances, and ηt is an (r × 1) vector of the disturbances driving the common factors. Ft and vt are taken to be mutually orthogonal due to their different natures, E(Ftνis) = 0 , and hence, E(ηjtνis) = 0 for all i, j, t, and s. Further assumptions are that vt are mutually orthogonal (i.e., E(vjtνis) = 0 for all i, j, t, s, i ≠ j) and that D(L) is a diagonal matrix.5 Substituting Eq. (2) into Eq. (1) and collecting terms, the dynamic factor model is written in VAR form. This is henceforth referred to as FAVAR:


Ft Xt




¼

ΦðLÞ ΛΦðLÞ

0 DðLÞ




 " F# F t−1 ε þ tX ; X t−1 εt

ð3Þ

where "

F

εt X εt

#


¼


  0 I ; ηt þ vt Λ

ð4Þ

4 (1) and (2) correspond to the static form representation of the dynamic factor model of Stock and Watson. In general, the static factor Ft can contain lags of the underlying dynamic factors. We assume for the applications at hand that the numbers of static and dynamic factors coincide. In this case, the static and dynamic factors are identical and there is no difference between the static and dynamic forms. See Stock and Watson for details. 5 With these assumptions, each variable in Xt is not affected by all other idiosyncratic disturbances than its own idiosyncratic disturbance across horizons. The diagonality of D(L) can be relaxed by introducing the lags of other variables to each equation in (1). The benefit is that the variable is allowed to be affected by other idiosyncratic disturbances at horizons other than the contemporaneous term. However, this comes at the cost of consuming degrees of freedom. The cost will be particularly severe when the model has a large set of variables with a short span of time series data, such as that in our study. The degree of freedom saved from the diagonality assumption of D(L) would also help accommodate enough lags of common factors (q) to take into full account the common interactions among variables, which is our main focus in this study.

452

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

with the covariance matrix E



0 εt εt

" ¼

Ση ΛΣη

# 0 Ση Λ ; 0 ΛΣη Λ þ Σv

ð5Þ

where Ση = E(ηtη′) t and Σv = E(vtvt′). Inverting the FAVAR representation in Eqs. (3) and (4), we can obtain the factor-augmented vector moving average (FAVMA) representation for Xt in terms of current and lagged values of ηt and vt: X t ¼ BðLÞηt þ C ðLÞvt ;

ð6Þ

where B(L) = (I − D(L)L)−1Λ(I − Φ(L)L)−1 and C(L) = (I − D(L)L)−1. 2.2. Estimation The FAVAR model in Eq. (3) is estimated by a sequential estimation strategy.6 The procedure first estimates the common factors, Ft, and the factor loadings, Λ, by solving the following minimization problem iteratively, min F 1 ; :::: F T ;Λ;DðLÞ T

−1

T X

0

½ðI−DðLÞLÞX t −Λ F t  ½ðI−DðLÞLÞX t −Λ F t ;

t¼1

where T denotes the sample size. A preliminary estimate of Ft is produced by the application of principal components analysis to Xt, and a preliminary estimate of D(L) is obtained by OLS estimation of Eq. (1). Then, the estimate Ft is updated by taking the principal components of the filtered variables (I − D(L)L)Xt, and conditional on the estimated common factors, an estimate of Λ and an updated estimate of D(L) are attained using OLS from Eq. (1). This iteration continues until convergence is achieved.7 The first step produces the estimates of Ft, Λ, and D(L). Second, given the final estimate of Ft, one can estimate Φ(L) by OLS from Eq. (2). The FAVAR model in Eq. (3) can be constructed using the final estimates of Ft, Λ, and D(L). In their procedure, Stock and Watson applied the principal component analysis to the entire set of variables in Xt with the number of common factors, r, chosen by the information criteria of Bai and Ng (2002). The common factors correspond to the r principal components associated with the r largest eigenvalues. This approach exploits all available information in the observed data, which is theoretically appealing. In practice, however, an economic interpretation of estimated common factors may be difficult, especially when the numbers of variables and factors are large. Bernanke and Boivin (2003), Boivin and Ng (2006) and Bagliano and Morana (2009) propose an alternative approach that can be more suitable for the studies in which the interpretability of factors is essential. The key feature is to divide the dataset into sub-sets of relatively homogenous variables and estimate the factors separately as the first principal component (i.e. having the largest eigenvalue) for each sub-set of variables. This method of extracting common factors from each group can make it easier to give economic meaning to the common factors. Separate estimation also avoids contamination from series potentially unrelated to the phenomenon of interest (see Boivin and Ng for an extensive treatment of this issue). A Monte Carlo study by Morana (2007) shows that principal components analysis is a very effective tool to extract common factors from a set of codependent variables even when the cross-sectional dimension is as low as two units. Our strategy follows the latter approach, but also takes advantage of a formal test to determine the number of common factors. Specifically, the variables in the model are grouped into eight blocks: (1) world oil prices, (2) world GDP, (3) regional GDP, (4) regional real exchange rates, (5) regional inflation, (6) regional money growth, (7) regional exports, and (8) regional imports. While a full discussion will be provided in Section 4, the Bai and Ng information criterion and principal component analysis indicate that blocks 3 through 8 share three common factors, namely, regional GDP factor, regional real exchange rate factor and regional nominal factor. Adding up the two world factors in blocks 1 and 2, we therefore posit that a total of five factors (r = 5) characterize co-movements among the variables: world oil price factor, world GDP factor, regional GDP factor, regional real exchange rate factor, and regional nominal factor. The regional GDP factor is estimated as the first principal component from block 3, and the regional real exchange rate factor is obtained in the same way from block 4. The regional nominal factor corresponds to the first principal component obtained from the merged set of blocks 5 and 6. No factors are assigned from blocks 7 and 8 because the variables in these blocks are shown to be explained largely by the aforementioned common factors. 2.3. Identifying structural shocks The FAVAR model entails structural identification to give economic interpretations of the factor disturbance ηt. As in typical structural VAR models, factor disturbances are related to the underlying structural shocks, denoted as ξt, in the following way: ξt ¼ Hηt :

6 7

It was proposed in detail by Stock and Watson (2005). In our empirical application, 10 iterations were sufficient to complete convergence.

ð7Þ

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

453

Here, H is an invertible (r × r) matrix, and the vector of the structural shocks ξt has a mean of zero and an identity covariance ma= Ir. The identification of structural shocks ξt amounts to an estimation of the elements in H with proper sets of restrictrix, i.e., E(ξt ξ′) t tions. For our model, no further identification is necessary for idiosyncratic disturbances vt because they are assumed to be mutually orthogonal in Eq. (1), i.e., Σv = E(vtvt′) is a diagonal matrix. Only the normalization of Σv is required, and this can be easily achieved from Σv = Θ−1(Θ−1)′, where Θ is an ((n × n)) diagonal matrix. The normalized idiosyncratic shock, denoted as ψt, is such that ψt = Θvt with a mean of zero and an identity covariance matrix, i.e., E(ψt ψ′t) = I .8 Note that identifying ξ in Eq. (7) is independent of n t identifying ψt according to the orthogonal condition between the two, i.e., E(ξt ψ′) = 0. Put altogether, the FAVMA = E(Hηtv′Θ′) t t representation in Eq. (6) can be rewritten in terms of the structural shocks as 



X t ¼ B ðLÞξt þ C ðLÞψt ;

ð8Þ

where B∗(L) = B(L)H−1 and C ∗(L) = C(L)Θ−1. Upon achieving identification, Eq. (8) can be utilized to examine how and to what extent the variables in Xt respond to the shocks ξt and ψt over time by means of the impulse response and variance decomposition analysis. For the identification of ξt in Eq. (7), we make use of block lower-triangular exclusion restrictions, as detailed in Stock and Watson (2005). This method estimates H by imposing Wold causal ordering on the blocks of variables through the relationship B∗(0) = ΛH−1 (∵ B(0) = Λ in Eq. (6)).9 To apply for our case of r = 5 and n = 62, reorganize Xt into five blocks of variables, each with mi elements, where m1+…+m5 = 62. The first and second blocks contain world oil prices and world GDP, respectively (m1 = m2 = 1) The GDPs and real exchange rates of 10 East Asian countries enter the third and fourth blocks, respectively (m3 = m4 = 10). The final block accommodates the remaining variables of inflation, money growth, exports and imports for the 10 countries (m5 = 40). Given these blocks of variables, B(0)* is specified to have the following block lower-triangular structure: 2



B0;11 6 B 6 0;21  6  Bð0Þ ¼ 6 B0;31 6 ð62x5Þ 4 ⋮  B0;51

0  B0;22  B0;32 ⋮  B0;52

0 0  B0;33 ⋮  B0;53

0 0 0  B0;44  B0;54

3 0 7 0 7 7 0 7 7 5 0  B0;55

ð9Þ

where B∗0,ij is an (mi × 1) vector measuring the responses of the variables in block i to structural shock j. There are assumed to be five structural shocks (ξt) governing the common factors, and they are a world oil price shock (j = 1), a world GDP shock ( j = 2), a regional GDP shock ( j = 3), a regional real exchange rate shock ( j = 4), and a regional nominal shock ( j = 5). Under the identification structure of Eq. (9), the structural shocks can have a contemporaneous effect on the variables in a lower block, but not allowed to affect the variables in an upper block contemporaneously. For example, the world oil price shock can have a contemporaneous effect on all the variables in blocks 2 through 5, but the world price of oil reacts only with a one-period lag to the other four structural shocks.10 The regional nominal shock ordered last is not allowed to affect the variables in all other blocks contemporaneously but only subsequently, whereas the variables in block 5 are affected contemporaneously by all of the structural shocks. Regarding the three regional structural shocks, the ordering is determined based upon the relative speeds of adjustments to shocks, as in Bernanke et al. (2005): the shocks related to relatively slow-moving variables (GDP and real exchange rates) come before the shocks originating from relatively fast-moving variables (nominal variables). The block lower triangular structure in Eq. (9) identifies H from B∗(0) = ΛH−1 and the orthogonal condition between structural shocks (i.e., E(ξt ξ't) = HΣη H'). ξt in Eq. (7) are then identified, and their effects on the variables can be obtained from Eq. (8). Stock and Watson provide a full account of the estimation procedure. 3. Data description The empirical procedure outlined above is undertaken for 10 major countries in East Asia: ASEAN5 (Indonesia [IN], Malaysia [MY], Singapore [SG], Thailand [TH], and the Philippines [PH]), China [CN], Hong Kong [HK], Taiwan [TW], Japan [JP], and Korea [KR]. These countries were selected based on their institutional and economic significance in the region. Except for Taiwan, they are also the signatories of the Chiang Mai Initiative. The ASEAN5 are the five original member countries of ASEAN, and all have seen substantial development through regional free-trade agreements and policy coordination efforts. The non-original member countries of ASEAN (now the ASEAN10) are not included in this study because they are not as economically significant and because the data are severely limited. China, Hong Kong, and Taiwan, constituting greater China, share several common features, such as language and cultural backgrounds, although they exhibit distinctive economic characteristics. Japan and Korea are two major industrialized countries in East Asia in close geographic and economic proximity to each other. 8 While the orthogonality between idiosyncratic disturbances is a standard assumption, it is possible to relax such that vt are correlated with each other and Σv is no longer a diagonal matrix. Then, additional n(n − 1)/2 restrictions are required to identify n2 parameters in Θ, apart from the orthogonal condition of E(ψt ψt′) = Θ∑vΘ′ = In. A difficulty is that economic theory does not provide the necessary information for modeling interactions between the idiosyncratic components of variables. The situation becomes practically implausible when the number of variables is large, as in our case of n = 62. 9 ⁎ Several alternative identification procedures are available, for example, based on restrictions on the long-run impact matrix B * (1) = ∑∞ i = 0Bi or restrictions on the

factor loading matrix Λ. See Stock and Watson for an excellent survey. 10 This can address the concern raised by Kilian (2009), allowing the world oil market to respond to economic conditions with a one period lag.

454

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

We use seasonally adjusted quarterly data for real GDP [GDP], real exchange rate [REX], CPI inflation [INF], M1 money growth [DMO], real exports [EXT], and real imports [IMT] for each of the 10 East Asian countries over the period of 1993:Q1 to 2010:Q4. A coherent set of quarterly data was not readily available prior to 1993 for many countries. Coincidentally, East Asian economies began to interact actively since the early 1990s. The real exchange rate is defined in terms of U.S. dollars for consistency across countries, and M1 money growth is used to capture movements in monetary policy and liquidity. To reflect the general conditions of the world economy, we utilize data on the world oil price (WR_OIL) and world GDP (WR_GDP). The former denotes the average of the U.K. Brent, Dubai, and West Texas Intermediate prices. The latter is obtained by aggregating the real GDPs of the U.S. and the EU; this accounts for about 50% of world GDP according to the 2010 IMF statistics. There are 62 variables altogether. All data were drawn from Global Insight except for real exports and real imports, which were drawn from the IMF's International Financial Statistics. All variables except CPI inflation and M1 money growth are in logarithms and are transformed into first differences.11 These stationary variables are normalized to have a mean of zero and unit variances to avoid size effects, as the variables with relatively large variance can dominate the factor estimates. The FAVAR model in Eq. (3) is estimated using the lag length of p = q = 3 for D(L) and Φ(L), together with a constant, and a time dummy for the period of 1997:Q3 to 1998:Q1 to take the Asian financial crisis into account. 12 4. World and regional common factors As discussed earlier, all variables under consideration are grouped into eight blocks according to their economic relevance: world oil price (block 1), world GDP (block 2), regional GDP (block 3), regional real exchange rates (block 4), regional inflation (block 5), regional money growth (block 6), regional exports (block 7), and regional imports (block 8). Blocks 3 through 8 contain 10 variables, respectively, corresponding to the 10 East Asian countries. We apply principal component analysis to each of these blocks, and the common factor is chosen as the first principal component having the largest eigenvalue. Table 1 reports the fractions of the variation explained by the four principal components, PCi for i = 1, 2, 3, and 4, where PC1 has the largest eigenvalue, PC2 has the second largest eigenvalue, and so on. The remaining principal components contribute little and are not reported. Starting from block 3 of the regional GDPs, the first principal component, PC1, accounts for 67% of the total variation and more than 50% of the variation in the individual GDPs of all countries. The second principal component, PC2, explains only an additional 7% of the total variation. Moving to block 4 of regional real exchange rates, more than 50% of the total variation is attributable to the first principal component, PC1, which also accounts for over 66% of the variation in the individual real exchange rates except in Japan, Hong Kong, and China. In Japan, PC1 accounts for 30% of the variation in the real exchange rate, while PC3 and PC4 make smaller but significant contributions. In Hong Kong and China, PC2 and PC3 are, respectively, the most influential determinants of real exchange rates. The failure of PC1 in these countries may be partly due to differences in the exchange rate system.13 These exceptions only pertain to specific countries, and the principal components other than PC1 contribute no significant gain in summarizing movements in the real exchange rate block as a whole. Blocks 5 and 6 contain inflation and money growth variables. These two blocks may be efficiently merged into one to reflect nominal movements in the region. To test this possibility, Table 2 reports the results obtained by applying the Bai and Ng (2002) information criterion. The ICp2 statistics select one common factor among 20 variables of inflation and money growth. When the two blocks are merged, Table 1 shows that the first principal component corresponding to the largest eigenvalue accounts for over 50% of the total variation and a large fraction of the variation in individual variables across countries, save for a few cases. As the other principal components do not contribute much, the first principal component is interpreted as the nominal factor that is responsible for common movements in blocks 5 and 6. We do not extract any common factor from the export and import blocks of 7 and 8. Economic theory suggests that exports are determined mainly by world GDP and real exchange rates, with imports determined mainly by domestic GDP and real exchange rates. This prediction is confirmed by the test results in Tables 1 and 2. The ICp2 information criterion chooses two common factors in each of two combined block groups: 2, 4, and 7 for the export function and 3, 4, and 8 for the import function. In the export function, the first and second principal components are mainly associated with movements in world GDP and the real exchange rate, while, in the import function, they are mainly associated with movements in domestic GDP and the real exchange rate. For exports, the sum of these two principal components explains more than 62% of the total variation and between 57 and 79% of the individual exports across countries. For imports, the sum of these two principal components explains more than 62% of the total variation and between 42 and 80% of the individual imports across countries. To sum up, we assume that the three common factors characterize the regional blocks of 3 through 8. The presence of three common factors is statistically confirmed by the Bai and Ng information criterion (ICp2 = −0.136) in Table 2. Together with the world oil price and world GDP, a total of five common factors are present in the model: the world oil price factor, the world GDP factor, the regional GDP factor, the regional real exchange rate factor, and the regional nominal factor. Fig. 1 illustrates how well the common and idiosyncratic components represent individual countries' GDPs, the key variable in this study, over time. It is apparent that the common components and GDP evolve closely in tandem for all countries, whereas the

11

The augmented Dickey–Fuller test confirmed that all variables were characterized as an I(1) process. The optimal lag length is based on the Akaike Information Criteria (AIC). The optimal lag length for the factors is found to be 3. The optimal lag length for individual equations varies; however, the optimal lag length for the majority of individual equations is found to be 3. So, we set the lag order p = q = 3. 13 Unlike the other countries, which adopt floating exchange rates, Hong Kong operates a fixed exchange rate regime by pegging its currency to the U.S. dollars. From 1995 to 2005, the Chinese currency, the yuan, was also pegged to the U.S. dollar. In July 2005, China announced a shift of its exchange rate regime to a managed float, tying the value of the yuan to a broad basket of foreign currencies. However, the yuan movement vis-à-vis the U.S. dollar has been relatively stable compared to other regional currencies. 12

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

455

Table 1 Principal component analysis results. PC1

PC2

PC3

PC4

PC1

PC2

PC3

PC4

0.69 0.07 0.74 0.08 0.08 0.02 0.05 0.11 0.03 0.00 0.27 0.02 0.04

0.56 0.06 0.80 0.00 0.00 0.03 0.01 0.00 0.06 0.44 0.01 0.01 0.00

0.51 0.05 0.85 0.00 0.01 0.04 0.03 0.18 0.00 0.00 0.00 0.00 0.25

Eigenval Prop Cum Prop JP_REX TH_REX MY_REX TW_REX SG_REX KR_REX PH_REX IN_REX HK_REX CN_REX

5.09 0.51 0.51 0.30 0.78 0.69 0.67 0.58 0.66 0.69 0.67 0.00 0.05

1.24 0.12 0.63 0.15 0.01 0.01 0.00 0.02 0.01 0.01 0.00 0.61 0.39

0.92 0.09 0.72 0.20 0.01 0.15 0.04 0.00 0.01 0.00 0.02 0.07 0.42

0.84 0.08 0.80 0.26 0.03 0.00 0.02 0.00 0.00 0.13 0.01 0.30 0.06

1.77 0.09 0.63 0.27 0.21 0.53 0.04 0.00 0.06 0.27 0.00 0.02 0.00

1.24 0.06 0.69 0.08 0.08 0.02 0.00 0.52 0.00 0.00 0.00 0.07 0.00

1.08 0.05 0.74 0.40 0.02 0.00 0.17 0.21 0.07 0.01 0.08 0.03 0.00

JP_DMO TH_DMO MY_DMO TW_DMO SG_DMO KR_DMO PH_DMO IN_DMO HK_DMO CN_DMO

0.07 0.87 0.90 0.56 0.77 0.66 0.78 0.82 0.89 0.84

0.02 0.03 0.04 0.08 0.06 0.06 0.02 0.00 0.05 0.03

0.39 0.00 0.00 0.00 0.00 0.03 0.03 0.00 0.00 0.01

0.02 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.03

1.43 0.07 0.69 0.01 0.09 0.00 0.00 0.00 0.10 0.01 0.00 0.02 0.45 0.42 0.02 0.02 0.02 0.00 0.05 0.01 0.03 0.05 0.00 0.12

1.03 0.05 0.74 0.00 0.26 0.02 0.10 0.09 0.02 0.00 0.01 0.00 0.06 0.30 0.00 0.00 0.11 0.01 0.03 0.01 0.00 0.00 0.00 0.00

Eigenval Prop Cum Prop JP_GDP TH_GDP MY_GDP TW_GDP SG_GDP KR_GDP PH_GDP IN_GDP HK_GDP CN_GDP JP_REX TH_REX MY_REX TW_REX SG_REX KR_REX PH_REX IN_REX HK_REX CN_REX JP_IMT TH_IMT MY_IMT TW_IMT SG_IMT KR_IMT PH_IMT IN_IMT HK_IMT CN_IMT

13.03 0.43 0.43 0.69 0.66 0.71 0.71 0.56 0.63 0.53 0.54 0.75 0.53 0.17 0.05 0.01 0.01 0.02 0.08 0.06 0.20 0.07 0.03 0.42 0.69 0.67 0.76 0.64 0.55 0.60 0.37 0.77 0.54

1.51 0.05 0.67 0.01 0.00 0.01 0.00 0.00 0.04 0.01 0.01 0.00 0.02 0.16 0.00 0.00 0.03 0.11 0.01 0.00 0.02 0.44 0.27 0.06 0.00 0.01 0.00 0.07 0.05 0.03 0.12 0.00 0.02

1.14 0.04 0.71 0.02 0.01 0.01 0.05 0.00 0.06 0.01 0.04 0.00 0.07 0.03 0.00 0.01 0.01 0.00 0.00 0.05 0.01 0.33 0.11 0.07 0.04 0.00 0.00 0.01 0.02 0.00 0.00 0.01 0.18

GDP Eigenval Prop Cum Prop JP_GDP TH_GDP MY_GDP TW_GDP SG_GDP KR_GDP PH_GDP IN_GDP HK_GDP CN_GDP

6.67 0.67 0.67 0.64 0.73 0.77 0.71 0.59 0.75 0.53 0.61 0.77 0.59

REX

INF + DMO Eigenval Prop Cum Prop JP_INF TH_INF MY_INF TW_INF SG_INF KR_INF PH_INF IN_INF HK_INF CN_INF

10.88 0.54 0.54 0.05 0.24 0.26 0.29 0.04 0.48 0.51 0.49 0.51 0.84

WR_GDP + REX + EXT Eigenval Prop Cum Prop WR_GDP JP_REX TH_REX MY_REX TW_REX SG_REX KR_REX PH_REX IN_REX HK_REX CN_REX JP_EXT TH_EXT MY_EXT TW_EXT SG_EXT KR_EXT PH_EXT IN_EXT HK_EXT CN_EXT

8.37 0.40 0.40 0.62 0.30 0.33 0.13 0.17 0.05 0.33 0.31 0.56 0.04 0.06 0.46 0.64 0.57 0.66 0.47 0.42 0.57 0.64 0.55 0.48

4.66 0.22 0.62 0.09 0.05 0.44 0.61 0.50 0.61 0.33 0.37 0.23 0.03 0.00 0.29 0.09 0.06 0.13 0.20 0.19 0.00 0.09 0.22 0.13

GDP + REX + IMT 5.69 0.19 0.62 0.01 0.04 0.06 0.00 0.02 0.09 0.00 0.00 0.02 0.00 0.19 0.72 0.75 0.63 0.63 0.58 0.65 0.58 0.00 0.02 0.00 0.03 0.05 0.00 0.00 0.25 0.10 0.24 0.01 0.00

Notes: In each panel, the first row reports the four largest eigenvalues (Eigenval) of the block. The second and third rows report the fractions (Prop) and cumulated fractions (Cum Prop) of the total variation in the block accounted for by each of the four principal components. The figures that follow are the fractions of the variation in the individual variables accounted for by each of the four principal components.

456

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

Table 2 Test of the number of common factors.

r=1

r=2

r=3

r=4

r=5

r=6

r=7

r=8

r=9

r = 10

INF+DMO 0.094

0.097

0.149

0.193

0.226

0.271

0.297

0.321

0.341

0.347

–0.325

–0.314

–0.279

–0.248

–0.234

–0.215

–0.192

–0.175

–0.171

–0.329

–0.293

–0.257

–0.232

–0.199

–0.175

–0.142

–0.108

–0.066

–0.136

–0.119

–0.085

–0.051

–0.009

0.031

0.070

0.118

WR_GDP+REX+EXT –0.059 GDP+REX+IMT –0.101

GDP+REX+INF+DMO+EXT+IMT 0.077

–0.083

Notes: Figures are ICp2 information criterion statistics developed by Bai and Ng (2002), and those that are shaded are the chosen numbers of common factors. The maximum number of common factors is set to r = 10.

idiosyncratic components play a minor role.14 This yields tentative evidence that GDP movements in the region are synchronized, lending support to the feasibility of a currency union. Fig. 2 reports the standard deviations of common and idiosyncratic components together with those of actual series across countries. In the model, there are two sources of heterogeneity between countries: asymmetric transmission of common shocks and idiosyncratic shocks. It is shown that the standard deviations of the common components exceed those of the idiosyncratic components in all variables. The observed heterogeneity is mainly attributable to the asymmetric transmission of common shocks rather than idiosyncratic shocks; the idiosyncratic components exhibit a low level of dispersion across countries over the sample period. Apparently, the evidence for East Asia is quite distinct from the experiences of euro countries. Based on an analogous FAVAR model, Eickmeier (2009) found that the idiosyncratic component accounted for a large portion of the movements in GDP and prices for a number of member countries over the period of 1980 to 2003. Furthermore, the dispersion of both GDP and prices across euro countries is mainly due to idiosyncratic shocks, while little else is explained by the asymmetric transmission of common shocks. 5. Structural analysis This section examines how and to what extent the identified structural shocks underlying the common components are transmitted to each country and cause cross-country heterogeneity by means of impulse responses and variance decompositions. 5.1. Variance decompositions Forecast error variance decompositions provide a way to assess the relative importance of structural shocks in accounting for variations in variables. Tables 3.1 through 3.5 present the forecast error variance decompositions at various horizons along with one-standard errors generated by 300 bootstrap replications. Starting from the top panel, the regional GDP shock is the most important determinant, accounting for more than 50% of the fluctuations in the GDPs across all countries at the contemporaneous horizon. As the forecasting horizon increases, the strength of the regional GDP shock attenuates and the world GDP shock gains importance instead, except in Indonesia. Both world and regional GDP shocks are equally important in accounting for the variation of GDPs at a horizon of 16 quarters. For Indonesia, the world GDP shock is unimportant across the horizons, and shocks to regional GDP and the real exchange rate explain most of the movements in GDP. The second panel shows that the real exchange rate shock accounts for the bulk of short-run variations in the real exchange rate. This shock continues to be important at long horizons, where the world and regional GDP shocks also account for a sizable portion of the variation in the rate. Two exceptions are Hong Kong and China. In these countries, the idiosyncratic shock contributes significantly to the forecast error variance of the real exchange rate, and its presence is visible even at long horizons. This may be partially due to their different exchange rate systems, as discussed in Section 4. At long horizons, the regional GDP shock is the most important determinant of the real exchange rate in Hong Kong and China. Moving to the third panel, the nominal shock is most important in explaining the forecast error variance of inflation at short horizons. The exception is Japan, where the shock exerts only marginal effects, and the same is true at long horizons. This is anticipated from the principal component analysis in Table 1, as the common nominal factor accounts for five percent of the variation in the Japanese inflation data. For the other countries, the world oil price shock is also important for short-term variation in inflation, but its importance diminishes as the forecasting horizon increases. At long horizons, the world and regional GDP shocks account for an increasing portion of the forecast error variance of inflation, while the nominal shock remains strong. The fourth panel reports the 14 While not reported due to space limitations, the common and idiosyncratic components of other regional variables have very similar implications. These results are available upon request.

JP_GDP

MY_GDP

0.75

0.6

0.50

0.4

SG_GDP

PH_GDP

0.4

HK_GDP

1.0

1.00

0.8

0.75

0.6

0.50

0.4

0.25

0.2

0.00

-0.0

-0.25

-0.2

-0.50

-0.4

-0.75

-0.6

-1.00

0.2 0.2

0.25

-0.0 -0.2

-0.25

-0.2 -0.4

-0.50

-0.6

-0.75

-0.8

-1.00

-1.0

-1.25

-1.2

-0.4 -0.6

1997

2000

2003

2006

2009

-0.8 1997

TH_GDP

2000

2003

2006

2009

-0.8 1997

TW_GDP 0.6

0.8

0.50

0.4

0.6

0.2

0.4

0.00

2003

2006

2009

-1.25 1997

2000

KR_GDP

0.75

0.25

2000

2003

2006

2009

IN_GDP

0.0

2009

0.50 0.25

-0.2 -0.4

0.00

-0.4

-1.00

-0.6

-1.25

-0.8

-0.8

-1.50

-1.0

-1.0

2009

2006

0.75

-0.5

-0.50

2006

2009

1.00

-0.2

2003

2006

1.25

-0.0

2000

2003

CN_GDP

1.0

0.2

-0.0

1997

2000

0.5

-0.25

-0.75

1997

-1.0

-0.25

-0.6

-0.50 -1.5

1997

2000

2003

2006

2009

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

0.00

-0.0

-0.75 -2.0 1997

2000

2003

2006

2009

-1.00 1997

2000

2003

2006

2009

1997

2000

2003

Fig. 1. Common and idiosyncratic components of individual countries' GDPs. Notes: The panels show GDP series (black), together with their common (blue) and idiosyncratic (red) components. All series are shown as eight-quarter centered moving averages for a better visuality.

457

458

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

GDP

INF

1.50

EXT 1.75

2.00

1.25 1.00

1.50 1.25

1.50

1.00 0.75

1.00

0.75

0.50

0.50

0.50

0.25

0.25

0.00

0.00 1995 1997 1999 2001 2003 2005 2007 2009

REX 2.00

0.00 1995 1997 1999 2001 2003 2005 2007 2009

1995 1997 1999 2001 2003 2005 2007 2009

DMO 2.5

IMT 2.00

1.75

1.75 2.0

1.50 1.25

1.50 1.25

1.5

1.00

1.00 1.0

0.75 0.50

0.75 0.50

0.5

0.25

0.25

0.00

0.0 1995 1997 1999 2001 2003 2005 2007 2009

0.00 1995 1997 1999 2001 2003 2005 2007 2009

1995 1997 1999 2001 2003 2005 2007 2009

Fig. 2. Dispersion of common and idiosyncratic components. Note: The panels show standard deviations of series (black), common components (blue), and idiosyncratic components (red).

variance decompositions for money growth, and the results vary across countries. While the nominal shock accounts for a sizable portion of the variation in money supply, the other shocks appear to be equally or more important in some countries. In particular, the idiosyncratic shock explains between 13 and 44% of the contemporaneous variation in money growth across countries. The results do not change much as the forecasting horizon increases. At long horizons, the regional GDP and nominal shocks explain a large portion of the variability in money growth for many countries. The idiosyncratic shock becomes less important, but it still has significant effects in several countries. The persistent presence of idiosyncratic shocks hints at the possibility that these countries exhibit some degree of heterogeneity in their monetary policies, a point we will revisit when discussing the impulse response analysis. The results for exports and imports in the final two panels display similar patterns. The world and regional GDP shocks play a major role in accounting for the short-term variation in exports and imports. The exceptions are that the real exchange rate shock is more important than the two shocks for exports in the Philippines and Indonesia and for imports in Japan, Thailand, and the Philippines, while the idiosyncratic shock is the most important determinant of Chinese exports and imports. In many countries, the nominal shock also explains a considerable fraction of the forecast error variance of both exports and imports. Prasad and Gable (1998) and Fisher and Huh (2002) similarly reported that nominal shocks are an important determinant of exports and imports in 23 developed countries and the trade balances in the G-7. As the forecasting horizon increases, the contribution of the nominal shock declines but is still significant in some countries, such as Indonesia and China. The world GDP shock increases in importance, becoming the major determinant of both exports and imports in most of the countries at a horizon of 16 quarters. The regional GDP shock remains significant, with strong evidence for imports in the Philippines and China. For the Philippines and China, the real exchange rate and nominal shocks account for 44 and 58% of the variation in their respective exports. 5.2. Impulse responses Figs. 3.1 to 3.5 display the responses of the series in levels to a one-unit shock in each structural disturbance.15 As shown in Fig. 3.1.1, GDP initially increases following a positive shock to the world oil price and begins to decrease after approximately two quarters, with a few exceptions. An initial increase in GDP is also reported by a group of VAR studies (e.g., Burbidge & Harrison, 1984; Hooker, 1996; Kilian, 2009), though this result is not consistent with standard economic theories. For the Philippines, China, and Indonesia, the responses remain positive over all horizons, and these oil-producing countries appear to capitalize on the rise in the oil price. Similar to the GDP response, exports and imports increase in the first few quarters following the shock and decline thereafter in all countries save Indonesia. The oil price shock causes inflation to rise immediately, while the effects are somewhat mixed as the forecast horizon increases. Fig. 3.2 shows the responses to a positive world GDP shock. All of the regional GDPs show 15 To conserve space as well as improve presentation, the confidence bands of the responses are not reported, but they are available upon request. In general, the effects of the world oil shock on the variables are statistically significant only at short horizons, except that the GDP responses are also significant at longer horizons. The world and regional GDP shocks produce responses that are mostly significant, while the significance is somewhat weaker for inflation in several countries. For the regional real exchange rate and regional nominal shocks, the significance of the responses is mixed depending on the variables and the countries. Nevertheless, GDP responds to these shocks significantly across countries.

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

459

Table 3.1 Forecast error variance decompositions at the contemporaneous horizon. WR_OIL JP_GDP TH_GDP MY_GDP TW_GDP SG_GDP KR_GDP PH_GDP IN_GDP HK_GDP CN_GDP JP_REX TH_REX MY_REX TW_REX SG_REX KR_REX PH_REX IN_REX HK_REX CN_REX JP_INF TH_INF MY_INF TW_INF SG_INF KR_INF PH_INF IN_INF HK_INF CN_INF JP_DMO TH_DMO MY_DMO TW_DMO SG_DMO KR_DMO PH_DMO IN_DMO HK_DMO CN_DMO JP_EXT TH_EXT MY_EXT TW_EXT SG_EXT KR_EXT PH_EXT IN_EXT HK_EXT CN_EXT JP_IMT TH_IMT MY_IMT TW_IMT SG_IMT KR_IMT PH_IMT IN_IMT HK_IMT CN_IMT

0.10 8.09 3.54 8.62 0.04 9.27 0.07 7.41 10.28 5.56 0.80 1.25 2.98 17.01 10.28 14.39 0.19 0.59 2.35 0.16 1.77 47.40 47.78 18.64 11.73 38.04 39.66 15.02 38.59 17.13 6.90 0.78 9.39 50.10 0.00 1.30 3.73 29.42 27.71 1.47 4.46 40.28 16.67 10.49 3.90 12.10 4.25 22.84 0.28 12.25 17.88 2.18 12.87 3.77 6.09 2.51 17.33 19.94 4.40 6.76

WR_GDP (14) (16) (11) (12) (13) (13) (16) (13) (12) (18) (20) (13) (10) (14) (10) (15) (11) (14) (17) (02) (15) (13) (17) (13) (13) (10) (13) (15) (11) (15) (13) (09) (13) (16) (15) (11) (10) (07) (10) (18) (15) (17) (15) (15) (17) (16) (12) (14) (13) (16) (08) (17) (17) (12) (13) (14) (20) (12) (15) (14)

33.32 31.60 12.65 24.26 36.90 22.18 33.33 0.25 18.83 31.00 1.99 11.27 9.29 7.00 4.24 23.00 11.50 18.24 27.96 3.43 42.43 4.43 13.57 1.38 18.01 0.29 13.11 4.35 0.26 16.45 2.99 3.11 24.09 0.27 20.19 7.65 0.42 0.92 2.17 2.93 56.00 11.28 27.35 36.27 29.38 27.99 11.10 0.16 28.07 24.04 4.37 27.16 34.28 56.33 25.68 17.27 1.57 1.12 18.01 0.36

GDP (10) (08) (04) (12) (08) (07) (15) (11) (08) (14) (20) (14) (14) (10) (09) (08) (12) (06) (11) (04) (17) (13) (07) (17) (07) (15) (08) (13) (15) (13) (07) (14) (13) (20) (11) (11) (19) (09) (12) (13) (08) (09) (09) (11) (12) (10) (18) (08) (08) (11) (16) (06) (10) (11) (07) (08) (11) (14) (08) (12)

58.49 49.55 70.88 60.96 59.53 58.31 60.01 68.45 61.38 59.13 1.51 4.57 5.88 0.11 1.86 7.95 7.61 7.39 48.43 8.25 17.15 9.46 0.21 5.73 25.11 0.01 1.75 2.88 0.79 17.23 8.50 24.85 12.99 0.29 18.91 19.44 52.25 11.64 31.13 6.49 25.00 4.21 13.00 31.53 49.15 32.55 5.44 0.05 48.78 14.93 6.97 0.06 17.04 28.17 32.98 40.00 23.51 8.44 58.10 17.33

(14) (18) (11) (14) (16) (14) (27) (23) (13) (19) (14) (15) (12) (15) (12) (12) (11) (15) (14) (05) (09) (14) (11) (17) (17) (08) (10) (12) (18) (12) (16) (16) (16) (20) (17) (11) (17) (11) (20) (14) (12) (19) (16) (13) (14) (14) (24) (15) (16) (19) (21) (17) (16) (16) (12) (15) (16) (14) (15) (17)

REX

NOM

IDS

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 89.87 77.59 73.93 71.75 80.22 47.41 73.92 64.58 0.74 23.36 20.22 5.75 0.43 12.80 7.60 0.01 8.00 5.21 0.14 10.10 29.37 4.34 0.17 9.39 32.75 0.01 3.88 3.15 0.03 6.65 0.09 1.11 4.14 0.20 1.80 4.32 37.87 24.76 0.49 1.51 29.79 33.00 0.01 3.70 2.14 11.94 31.65 9.05 1.20 0.36

0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 4.73 30.34 32.50 50.11 15.90 55.28 34.43 66.29 57.18 33.59 13.93 22.85 37.48 0.70 6.45 28.43 23.53 42.06 26.47 67.76 5.50 37.23 33.55 17.35 7.07 19.50 9.42 44.96 1.87 20.50 31.29 0.82 28.45 2.34 20.18 17.64 14.13 50.91 2.33 24.09

8.09 10.76 12.93 6.17 3.54 10.24 6.60 23.89 9.51 4.31 5.82 5.33 7.93 4.13 3.40 7.25 6.79 9.21 20.51 64.80 13.69 2.61 5.50 11.34 21.64 6.38 3.05 6.25 3.04 5.49 38.32 44.06 15.89 39.24 21.70 43.18 16.19 12.80 12.50 14.70 8.94 5.89 5.30 4.16 8.69 3.54 31.92 7.23 20.51 26.78 9.70 36.78 7.34 5.69 12.94 10.64 11.81 10.54 15.96 51.11

(24) (16) (14) (20) (16) (12) (13) (15) (18) (05) (16) (07) (08) (13) (14) (14) (17) (13) (15) (15) (14) (14) (07) (14) (08) (17) (14) (11) (15) (07) (06) (12) (08) (04) (07) (08) (09) (21) (06) (13) (08) (06) (06) (06) (04) (05) (08) (22) (06) (18)

(23) (25) (23) (18) (16) (23) (22) (20) (24) (23) (25) (20) (19) (22) (22) (25) (22) (22) (25) (20) (05) (07) (06) (04) (05) (06) (20) (17) (06) (10) (10) (10) (06) (04) (03) (08) (09) (14) (05) (09)

(02) (06) (01) (02) (04) (03) (18) (11) (03) (09) (12) (09) (05) (07) (04) (02) (06) (05) (16) (08) (08) (06) (03) (07) (05) (12) (05) (12) (06) (14) (13) (12) (06) (09) (11) (13) (09) (09) (13) (07) (03) (04) (07) (02) (07) (02) (07) (14) (02) (06) (07) (04) (04) (02) (04) (02) (08) (18) (03) (15)

Notes: Figures in parentheses are one-standard errors computed using 300 bootstrap replications of the model. ‘IDS’ in the last column denotes the idiosyncratic shock.

an increase, and the responses are similarly hump-shaped with the exception of Indonesia. A positive world GDP shock leads to higher inflation. As the world GDP increases, exports in the region increase, with the effects particularly pronounced in Korea, Japan, and Singapore. Imports also increase due to income effects, except in China, which shows negative responses after five quarters. Fig. 3.3 shows GDP increases in all countries when there is a positive regional GDP shock. Similar to the responses to a world GDP shock, responses are hump-shaped and synchronized between countries. In the variance decomposition analysis, the regional GDP shock was the main cause of short-run movements in GDP, while both world and regional GDP shocks were equally important at

460

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

Table 3.2 Forecast error variance decompositions at the one-quarter horizon. WR_OIL JP_GDP TH_GDP MY_GDP TW_GDP SG_GDP KR_GDP PH_GDP IN_GDP HK_GDP CN_GDP JP_REX TH_REX MY_REX TW_REX SG_REX KR_REX PH_REX IN_REX HK_REX CN_REX JP_INF TH_INF MY_INF TW_INF SG_INF KR_INF PH_INF IN_INF HK_INF CN_INF JP_DMO TH_DMO MY_DMO TW_DMO SG_DMO KR_DMO PH_DMO IN_DMO HK_DMO CN_DMO JP_EXT TH_EXT MY_EXT TW_EXT SG_EXT KR_EXT PH_EXT IN_EXT HK_EXT CN_EXT JP_IMT TH_IMT MY_IMT TW_IMT SG_IMT KR_IMT PH_IMT IN_IMT HK_IMT CN_IMT

9.07 8.01 7.25 3.64 0.42 4.48 10.26 2.33 5.49 3.46 1.17 1.23 4.28 20.66 12.61 12.75 0.18 1.76 9.11 0.28 10.48 38.12 48.15 21.39 12.08 30.66 38.65 19.40 36.02 13.47 30.79 1.79 8.17 18.79 8.73 0.94 3.46 21.27 32.35 5.60 19.10 28.81 20.01 11.98 6.24 7.67 6.19 24.05 13.81 5.49 9.16 26.23 14.47 13.48 9.82 7.08 25.11 24.91 15.55 7.58

WR_GDP (10) (15) (11) (11) (13) (10) (18) (10) (11) (16) (16) (14) (10) (16) (11) (14) (11) (13) (15) (04) (17) (12) (12) (17) (12) (16) (09) (16) (12) (15) (14) (09) (13) (15) (15) (10) (13) (09) (12) (17) (09) (13) (14) (12) (14) (12) (15) (12) (11) (14) (12) (14) (15) (12) (11) (13) (16) (15) (13) (16)

39.02 33.29 38.96 44.83 48.56 41.87 29.55 1.06 58.17 35.25 1.01 24.34 15.88 14.70 9.83 28.49 17.85 28.33 17.05 8.83 42.94 5.13 12.96 11.83 41.02 1.52 7.72 3.65 1.73 23.28 11.33 2.91 18.91 31.11 18.83 5.26 11.77 4.33 6.11 5.01 46.00 25.62 32.57 38.35 39.29 46.09 37.52 0.25 47.34 52.68 40.31 36.23 41.67 43.97 38.08 39.97 17.74 3.96 38.27 15.50

GDP (11) (12) (06) (10) (09) (10) (15) (05) (08) (13) (16) (14) (15) (12) (08) (10) (13) (09) (10) (04) (18) (12) (10) (15) (08) (14) (11) (10) (15) (13) (11) (13) (14) (13) (11) (08) (15) (08) (14) (18) (09) (09) (14) (10) (13) (14) (17) (05) (07) (10) (10) (05) (09) (10) (09) (13) (11) (10) (06) (11)

44.20 49.89 40.73 41.83 43.85 41.64 43.84 44.95 29.16 56.40 10.08 4.34 21.10 0.38 7.60 6.79 15.41 6.93 42.37 14.53 19.32 15.96 2.39 8.30 16.88 2.16 7.95 5.79 7.65 22.64 11.07 30.62 13.34 27.04 14.80 16.47 43.45 14.32 12.07 7.62 18.62 5.36 17.33 23.91 32.75 22.74 6.91 0.10 24.23 12.89 10.45 3.15 22.04 24.55 22.40 22.26 26.73 8.68 25.29 8.91

REX (12) (18) (13) (14) (15) (14) (24) (18) (12) (18) (12) (14) (12) (18) (13) (12) (09) (13) (13) (05) (10) (16) (13) (15) (12) (10) (07) (10) (16) (11) (16) (12) (14) (18) (14) (11) (15) (14) (23) (14) (10) (13) (16) (12) (12) (13) (20) (13) (11) (14) (15) (12) (16) (15) (12) (14) (14) (19) (13) (16)

0.64 1.57 6.79 5.10 0.89 7.28 5.57 22.41 0.41 0.17 79.31 60.87 53.67 59.18 63.70 43.37 57.16 54.81 1.73 24.38 12.75 8.15 2.95 9.07 8.53 18.61 6.25 3.33 1.70 6.82 15.64 3.57 0.25 2.75 27.06 40.63 3.86 17.47 6.89 6.07 1.40 4.16 2.21 4.95 2.24 2.76 15.15 13.09 0.81 2.41 18.43 11.34 1.71 5.32 3.48 23.70 12.36 6.88 0.88 6.31

NOM (01) (03) (01) (02) (01) (01) (07) (08) (01) (03) (21) (15) (15) (19) (16) (13) (14) (15) (17) (06) (14) (06) (08) (12) (09) (12) (15) (10) (11) (11) (10) (11) (07) (08) (10) (12) (13) (10) (10) (08) (04) (08) (05) (06) (05) (05) (10) (16) (04) (06) (05) (04) (04) (05) (03) (04) (07) (10) (03) (11)

4.56 0.88 2.41 1.76 4.74 0.42 7.18 17.73 2.92 1.12 2.15 4.03 0.58 3.19 1.36 5.29 2.96 0.00 1.79 1.27 5.74 30.18 31.64 41.57 16.87 41.81 37.48 65.87 50.42 27.48 8.10 25.34 46.67 15.04 18.26 20.78 22.77 35.80 37.93 60.71 13.16 33.31 24.11 19.58 16.91 19.26 21.65 56.66 11.05 6.42 14.05 18.91 17.75 10.91 22.29 4.21 11.23 52.26 17.72 8.51

IDS (02) (03) (01) (02) (03) (05) (04) (06) (04) (09) (06) (04) (01) (03) (02) (03) (03) (02) (08) (01) (23) (22) (22) (16) (06) (21) (22) (21) (23) (19) (22) (20) (17) (20) (20) (23) (16) (21) (24) (16) (04) (03) (06) (04) (05) (03) (13) (18) (05) (03) (05) (04) (04) (05) (03) (05) (07) (10) (05) (08)

2.51 6.37 3.87 2.84 1.54 4.31 3.60 11.51 3.86 3.60 6.28 5.19 4.49 1.89 4.90 3.32 6.44 8.16 27.95 50.71 8.78 2.45 1.90 7.84 4.61 5.24 1.94 1.96 2.47 6.31 23.07 35.77 12.66 5.27 12.32 15.92 14.69 6.81 4.66 14.98 1.72 2.74 3.77 1.22 2.56 1.47 12.59 5.86 2.75 20.12 7.60 4.13 2.36 1.77 3.92 2.80 6.83 3.32 2.28 53.19

(01) (04) (01) (02) (02) (02) (13) (04) (01) (05) (08) (08) (07) (04) (05) (02) (04) (05) (12) (10) (04) (08) (04) (07) (02) (06) (04) (09) (04) (09) (06) (07) (04) (04) (07) (07) (05) (06) (06) (05) (01) (02) (02) (01) (02) (01) (05) (08) (01) (02) (01) (01) (01) (01) (01) (01) (03) (06) (01) (06)

Notes: Figures in parentheses are one-standard errors computed using 300 bootstrap replications of the model. ‘IDS’ in the last column denotes the idiosyncratic shock.

long horizons. Taken together, the implication is that business cycle fluctuations in the region are synchronized. Exports and imports increase in a similar manner following a regional GDP shock. The exception is China, where both exports and imports decline after an initial increase in the first few quarters. Similar results were found in the Chinese response to a world GDP shock. Fig. 3.4 reports the effects of a shock leading to depreciation in the real exchange rate. Exports exhibit J-curve responses, increasing after an initial decline. The exceptions are the Philippines and Indonesia, where exports never decrease. GDP also declines initially, but

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

461

Table 3.3 Forecast error variance decompositions at the four-quarter horizon. WR_OIL JP_GDP TH_GDP MY_GDP TW_GDP SG_GDP KR_GDP PH_GDP IN_GDP HK_GDP CN_GDP JP_REX TH_REX MY_REX TW_REX SG_REX KR_REX PH_REX IN_REX HK_REX CN_REX JP_INF TH_INF MY_INF TW_INF SG_INF KR_INF PH_INF IN_INF HK_INF CN_INF JP_DMO TH_DMO MY_DMO TW_DMO SG_DMO KR_DMO PH_DMO IN_DMO HK_DMO CN_DMO JP_EXT TH_EXT MY_EXT TW_EXT SG_EXT KR_EXT PH_EXT IN_EXT HK_EXT CN_EXT JP_IMT TH_IMT MY_IMT TW_IMT SG_IMT KR_IMT PH_IMT IN_IMT HK_IMT CN_IMT

3.33 2.58 5.29 5.28 2.61 6.56 9.30 0.67 7.72 1.27 6.11 0.72 2.27 15.76 4.73 7.36 1.85 1.11 15.57 0.39 3.41 26.18 34.13 17.93 6.34 21.39 23.09 16.07 23.89 11.46 25.76 15.93 24.78 12.36 8.60 9.31 4.86 18.95 26.09 30.73 6.35 13.95 7.84 5.16 2.21 3.36 12.23 12.75 5.23 10.49 5.41 11.88 6.17 5.79 2.87 4.31 4.84 14.87 6.29 5.47

WR_GDP (13) (17) (17) (15) (16) (12) (17) (12) (17) (16) (16) (14) (09) (16) (10) (18) (10) (14) (12) (06) (16) (12) (12) (14) (16) (14) (08) (15) (13) (16) (10) (17) (15) (13) (11) (09) (18) (12) (14) (18) (14) (16) (15) (16) (15) (15) (19) (15) (15) (19) (15) (14) (17) (16) (15) (16) (18) (19) (15) (20)

44.73 37.85 44.80 49.81 52.49 42.93 30.35 9.31 56.54 30.41 0.77 35.64 23.68 22.51 20.55 37.59 23.52 23.79 8.93 19.05 41.59 7.89 19.06 20.08 49.97 5.03 15.58 2.37 7.09 24.19 9.72 11.36 7.75 38.43 20.85 6.87 10.45 3.87 13.78 4.15 53.60 47.27 52.70 50.34 52.86 62.29 29.27 11.17 58.25 29.69 62.75 51.20 55.96 52.99 58.41 47.34 33.26 20.11 51.14 19.71

GDP (14) (13) (10) (10) (11) (11) (11) (04) (10) (14) (16) (10) (14) (12) (10) (15) (15) (10) (14) (04) (16) (15) (09) (14) (10) (09) (10) (09) (17) (16) (14) (14) (14) (12) (12) (13) (17) (08) (10) (20) (11) (12) (17) (14) (12) (16) (19) (08) (09) (17) (10) (09) (12) (10) (12) (16) (11) (11) (07) (13)

46.30 52.19 37.47 39.34 40.07 36.56 42.07 46.01 22.34 63.47 17.58 8.67 33.72 2.46 22.15 15.52 24.71 10.69 23.21 22.53 43.10 29.90 13.89 9.44 28.12 23.22 17.96 9.60 23.24 29.48 20.68 10.98 11.43 34.49 19.79 28.41 45.65 11.19 20.52 8.84 32.39 13.59 27.33 31.22 34.48 23.27 9.19 12.20 29.05 7.40 22.23 11.12 27.33 32.83 25.50 24.41 49.36 21.83 32.43 4.92

REX (12) (18) (16) (18) (18) (15) (19) (18) (18) (21) (16) (11) (12) (15) (17) (13) (09) (10) (14) (11) (11) (15) (17) (14) (14) (13) (10) (12) (16) (15) (17) (08) (12) (20) (12) (13) (16) (14) (21) (17) (13) (11) (17) (12) (13) (14) (20) (16) (16) (13) (16) (15) (16) (17) (13) (15) (14) (20) (15) (15)

2.37 1.89 4.54 2.04 0.99 3.88 8.43 32.43 0.70 0.67 71.65 43.21 36.54 56.21 45.63 30.47 40.31 54.74 6.82 22.68 3.97 9.33 1.59 7.91 4.94 12.16 3.91 6.53 3.30 4.97 16.06 8.39 4.39 1.43 25.24 29.84 2.53 30.07 6.09 9.69 1.48 3.75 1.00 1.99 1.60 2.26 36.24 15.44 1.34 5.02 3.94 13.96 1.88 2.30 3.39 12.54 7.13 2.74 1.49 2.80

NOM (02) (06) (02) (04) (03) (04) (05) (11) (02) (05) (17) (15) (15) (19) (18) (12) (16) (15) (15) (06) (13) (10) (07) (08) (10) (11) (13) (07) (09) (11) (09) (10) (08) (08) (11) (13) (13) (11) (10) (08) (04) (05) (06) (05) (05) (08) (09) (14) (04) (03) (03) (05) (05) (05) (03) (05) (06) (03) (03) (11)

2.46 3.00 6.56 2.50 3.09 7.01 7.83 7.72 9.90 2.30 0.92 6.61 1.15 1.87 3.38 3.78 2.45 4.63 7.43 5.08 4.72 23.52 29.80 36.35 9.14 32.32 38.04 64.15 40.01 19.97 6.51 42.78 46.10 10.70 12.08 14.36 22.85 29.99 30.72 36.86 5.47 19.69 8.68 10.35 7.54 7.81 8.35 44.45 4.52 37.20 3.51 9.32 7.18 4.82 8.12 10.09 2.65 38.95 7.07 18.55

IDS (05) (05) (04) (08) (04) (06) (04) (06) (08) (11) (13) (08) (04) (09) (06) (06) (03) (03) (15) (03) (22) (21) (20) (16) (10) (15) (19) (21) (22) (16) (18) (17) (18) (20) (17) (17) (15) (21) (18) (16) (05) (04) (08) (06) (05) (05) (12) (16) (06) (05) (06) (04) (05) (07) (04) (07) (06) (10) (05) (12)

0.81 2.49 1.35 1.04 0.76 3.06 2.02 3.86 2.80 1.88 2.97 5.15 2.65 1.18 3.55 5.28 7.16 5.05 38.04 30.27 3.22 3.19 1.53 8.30 1.48 5.88 1.43 1.28 2.46 9.94 21.26 10.55 5.54 2.58 13.43 11.20 13.66 5.93 2.80 9.73 0.71 1.76 2.44 0.94 1.31 1.00 4.72 3.99 1.61 10.20 2.16 2.52 1.49 1.27 1.71 1.32 2.76 1.50 1.57 48.55

(01) (03) (01) (02) (02) (06) (01) (02) (01) (04) (03) (08) (08) (02) (02) (03) (04) (03) (11) (13) (03) (05) (02) (04) (01) (04) (05) (04) (02) (07) (04) (05) (03) (02) (05) (04) (03) (04) (04) (04) (01) (01) (02) (01) (01) (02) (03) (03) (01) (01) (01) (01) (01) (01) (01) (01) (01) (01) (01) (04)

Notes: Figures in parentheses are one-standard errors computed using 300 bootstrap replications of the model. ‘IDS’ in the last column denotes the idiosyncratic shock.

the responses afterward differ across countries. Interestingly, GDPs increase in developed economies and decrease in developing economies.16 Real depreciation causes imports to decline at short horizons in most countries. As the time horizon increases, the responses vary depending on the country, but many are not very significantly different from zero (not shown). Finally, Fig. 3.5 shows 16 One reason may be that the domestic industry in developed economies expands the capacities to exploit the depreciated real change rate (e.g. promoting cheaper exports and substituting dearer imports). Due to capacity constraints, this option may not be available for developing economies.

462

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

Table 3.4 Forecast error variance decompositions at the eight-quarter horizon. WR_OIL JP_GDP TH_GDP MY_GDP TW_GDP SG_GDP KR_GDP PH_GDP IN_GDP HK_GDP CN_GDP JP_REX TH_REX MY_REX TW_REX SG_REX KR_REX PH_REX IN_REX HK_REX CN_REX JP_INF TH_INF MY_INF TW_INF SG_INF KR_INF PH_INF IN_INF HK_INF CN_INF JP_DMO TH_DMO MY_DMO TW_DMO SG_DMO KR_DMO PH_DMO IN_DMO HK_DMO CN_DMO JP_EXT TH_EXT MY_EXT TW_EXT SG_EXT KR_EXT PH_EXT IN_EXT HK_EXT CN_EXT JP_IMT TH_IMT MY_IMT TW_IMT SG_IMT KR_IMT PH_IMT IN_IMT HK_IMT CN_IMT

7.32 2.31 7.66 8.18 4.27 9.13 7.40 0.65 12.04 0.97 11.17 0.54 1.79 15.01 4.11 6.98 1.40 1.95 13.54 0.26 2.69 25.67 33.58 15.62 5.27 19.39 20.73 13.81 20.47 10.66 21.49 11.88 21.35 8.37 7.88 12.45 4.39 18.90 23.47 30.50 5.92 10.47 7.09 4.27 2.61 2.74 20.54 8.23 6.82 19.90 9.50 9.41 5.03 5.51 3.17 5.52 4.08 8.65 6.98 7.63

WR_GDP (14) (17) (19) (16) (16) (13) (17) (12) (19) (16) (16) (15) (09) (16) (10) (18) (10) (15) (12) (08) (16) (13) (12) (15) (17) (15) (08) (16) (14) (15) (09) (18) (15) (14) (13) (09) (17) (15) (14) (19) (16) (18) (15) (18) (16) (17) (20) (18) (17) (21) (18) (15) (20) (18) (17) (18) (20) (21) (16) (21)

41.22 36.14 43.87 48.27 54.32 41.36 30.90 7.33 52.39 26.04 0.58 36.20 23.20 24.75 20.24 36.68 25.30 19.03 12.48 25.89 42.19 6.15 18.83 19.33 54.45 4.45 17.18 3.96 7.72 22.89 9.37 15.54 11.24 44.74 18.02 6.28 12.72 3.15 18.24 4.31 56.97 58.16 56.30 54.58 58.29 68.50 17.90 24.54 60.23 7.20 59.53 58.37 62.19 56.37 64.62 47.41 33.27 30.32 55.62 8.67

GDP (16) (13) (12) (12) (13) (12) (13) (05) (12) (20) (16) (09) (14) (14) (12) (17) (18) (14) (14) (05) (16) (17) (10) (15) (12) (09) (11) (11) (18) (16) (18) (18) (15) (12) (12) (18) (20) (08) (12) (23) (12) (14) (18) (15) (14) (18) (22) (09) (11) (19) (11) (11) (15) (11) (15) (19) (13) (13) (07) (14)

38.70 51.73 32.57 35.38 37.68 29.67 42.20 44.28 15.37 62.56 16.61 10.00 37.01 2.11 24.55 13.39 29.29 8.30 31.72 28.96 46.31 27.09 13.71 7.83 27.43 31.24 19.01 13.73 28.45 27.82 17.36 9.91 13.67 35.94 25.60 27.63 44.45 10.43 21.62 11.24 29.97 11.54 21.78 30.00 32.19 21.71 4.36 18.25 23.73 3.21 18.66 8.65 23.75 30.97 23.38 20.46 47.59 28.23 28.13 30.85

REX (13) (19) (17) (19) (21) (15) (19) (19) (20) (23) (17) (11) (12) (12) (18) (13) (12) (10) (14) (14) (12) (16) (19) (14) (17) (14) (12) (12) (15) (16) (20) (08) (12) (21) (15) (15) (17) (15) (22) (17) (14) (13) (16) (13) (14) (15) (21) (18) (17) (13) (18) (15) (17) (18) (14) (15) (15) (21) (15) (15)

4.92 1.66 3.43 2.72 0.79 3.24 10.93 40.67 1.25 2.73 67.07 38.23 34.37 55.51 41.59 28.56 32.69 60.51 3.71 19.08 2.67 13.62 1.85 9.81 3.77 10.15 4.11 4.06 3.03 4.47 14.62 13.86 7.07 1.15 21.83 26.58 4.86 28.38 5.80 11.91 2.03 2.86 3.18 1.60 1.05 1.49 38.59 12.96 0.96 7.95 4.81 11.93 1.54 1.68 2.26 8.34 8.78 2.57 1.36 0.45

NOM (02) (07) (02) (06) (04) (05) (06) (13) (02) (05) (17) (15) (16) (19) (19) (12) (17) (15) (13) (07) (14) (10) (08) (08) (11) (11) (12) (13) (07) (11) (10) (10) (09) (10) (11) (14) (13) (12) (10) (10) (05) (06) (07) (06) (06) (09) (09) (12) (05) (02) (04) (06) (06) (06) (04) (05) (08) (03) (04) (12)

7.12 5.48 11.12 4.36 2.25 12.85 6.75 4.04 16.44 6.08 2.02 9.73 1.11 1.50 6.79 5.93 4.02 3.96 3.06 6.13 3.15 23.36 30.38 38.23 7.44 27.38 37.49 63.21 37.54 18.57 12.50 39.98 41.61 7.17 11.10 13.72 18.35 32.50 28.14 33.88 4.36 14.75 8.88 8.35 4.50 4.59 15.81 31.90 6.65 56.90 5.94 8.52 5.77 3.84 4.89 16.94 3.73 28.94 6.10 30.15

IDS (07) (08) (05) (10) (06) (08) (04) (05) (09) (10) (14) (10) (05) (12) (08) (09) (05) (03) (17) (04) (23) (22) (21) (18) (12) (15) (20) (20) (23) (16) (19) (16) (18) (19) (19) (18) (16) (21) (18) (17) (06) (07) (10) (07) (06) (06) (13) (14) (08) (07) (08) (05) (05) (07) (06) (09) (05) (11) (06) (15)

0.72 2.69 1.34 1.09 0.69 3.76 1.83 3.03 2.51 1.63 2.56 5.29 2.52 1.12 2.72 8.47 7.30 6.25 35.49 19.67 2.99 4.10 1.65 9.18 1.64 7.40 1.47 1.22 2.80 15.60 24.66 8.84 5.06 2.63 15.57 13.33 15.23 6.62 2.73 8.15 0.75 2.22 2.77 1.20 1.36 0.97 2.81 4.13 1.61 4.83 1.56 3.12 1.72 1.64 1.68 1.34 2.56 1.29 1.81 22.25

(01) (04) (01) (03) (02) (07) (01) (02) (01) (05) (03) (08) (08) (02) (01) (05) (04) (04) (09) (14) (03) (04) (02) (05) (01) (04) (06) (03) (03) (07) (04) (05) (03) (03) (04) (04) (04) (05) (04) (05) (01) (01) (02) (01) (01) (02) (04) (04) (01) (01) (01) (01) (01) (01) (01) (02) (01) (01) (01) (05)

Notes: Figures in parentheses are one-standard errors computed using 300 bootstrap replications of the model. ‘IDS’ in the last column denotes the idiosyncratic shock.

that a positive nominal shock leads to higher inflation in all countries. The GDP increases initially, but the effects are short-lived. That nominal shocks have only transitory effects on GDP is documented in many studies (e.g., Eichenbaum & Evans, 1995). The exceptions are China and the Philippines, where the nominal shock continues to increase the GDPs. Exports increase initially due to the accompanied depreciation of real exchange rates. As the real depreciation is transitory (e.g., purchasing power parity), the increase in exports is lessened as the forecasting horizon increases. Imports initially increase for most countries, and this suggests that the

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

463

Table 3.5 Forecast error variance decompositions at the 16-quarter horizon. WR_OIL JP_GDP TH_GDP MY_GDP TW_GDP SG_GDP KR_GDP PH_GDP IN_GDP HK_GDP CN_GDP JP_REX TH_REX MY_REX TW_REX SG_REX KR_REX PH_REX IN_REX HK_REX CN_REX JP_INF TH_INF MY_INF TW_INF SG_INF KR_INF PH_INF IN_INF HK_INF CN_INF JP_DMO TH_DMO MY_DMO TW_DMO SG_DMO KR_DMO PH_DMO IN_DMO HK_DMO CN_DMO JP_EXT TH_EXT MY_EXT TW_EXT SG_EXT KR_EXT PH_EXT IN_EXT HK_EXT CN_EXT JP_IMT TH_IMT MY_IMT TW_IMT SG_IMT KR_IMT PH_IMT IN_IMT HK_IMT CN_IMT

8.24 1.34 7.52 8.76 4.72 11.28 7.96 1.76 13.76 1.46 12.69 0.42 2.08 14.46 4.52 5.88 1.19 1.89 5.50 0.21 1.57 24.26 33.09 12.50 3.39 18.05 19.35 14.20 17.25 8.77 20.41 12.29 21.29 5.88 6.17 11.69 3.37 19.55 21.80 30.91 4.40 6.78 6.03 2.71 2.32 3.04 20.99 4.88 6.42 18.73 10.21 6.48 3.67 3.76 3.00 6.40 2.72 6.45 5.44 4.01

WR_GDP (15) (16) (21) (17) (16) (14) (17) (12) (20) (16) (17) (16) (09) (16) (10) (18) (11) (16) (13) (09) (17) (15) (13) (15) (18) (16) (09) (17) (14) (15) (09) (20) (17) (15) (14) (09) (18) (17) (14) (20) (17) (20) (15) (20) (17) (18) (20) (19) (17) (21) (19) (16) (21) (19) (17) (19) (21) (23) (17) (20)

38.84 35.50 44.61 48.04 55.27 38.54 28.83 5.30 49.08 21.30 0.32 35.53 21.75 26.45 17.60 37.81 26.34 15.25 11.63 27.16 42.84 4.76 17.25 19.77 57.22 3.32 17.76 3.61 7.86 22.46 9.11 15.31 10.30 47.84 15.38 6.13 18.05 2.59 20.61 2.77 59.40 68.18 56.53 58.38 61.70 71.46 13.10 29.37 62.36 1.95 58.08 67.26 66.52 59.70 68.65 46.85 32.78 32.95 60.34 11.91

GDP (17) (13) (14) (15) (16) (14) (15) (08) (14) (22) (17) (08) (16) (15) (13) (18) (21) (16) (17) (06) (16) (19) (12) (18) (13) (09) (13) (14) (19) (17) (20) (20) (17) (12) (13) (21) (22) (09) (14) (25) (13) (16) (21) (16) (15) (19) (24) (11) (13) (21) (13) (13) (18) (13) (16) (22) (15) (14) (08) (15)

33.88 53.78 31.93 33.68 37.34 23.52 40.76 40.44 12.33 58.32 13.92 11.09 37.19 2.08 24.26 10.95 31.53 5.82 38.27 31.29 48.54 26.17 13.19 6.08 28.80 40.68 19.45 14.29 33.12 27.48 13.91 8.81 13.67 37.96 29.54 32.99 43.24 10.32 22.58 9.73 29.18 9.94 16.98 29.78 31.28 20.83 2.02 20.58 21.34 4.48 17.24 6.68 21.42 31.11 22.19 17.57 46.97 30.44 27.20 42.98

REX (14) (19) (18) (20) (22) (16) (21) (20) (21) (24) (18) (12) (12) (12) (18) (15) (13) (11) (14) (15) (13) (18) (20) (14) (19) (15) (13) (13) (15) (16) (21) (09) (12) (22) (16) (17) (18) (16) (22) (18) (15) (14) (16) (15) (15) (16) (22) (19) (17) (12) (19) (15) (18) (19) (15) (15) (16) (22) (15) (15)

8.91 2.18 2.34 3.60 0.45 3.86 12.79 48.05 2.95 6.50 68.37 36.20 34.96 54.79 41.48 24.15 29.45 67.24 13.29 18.76 1.69 16.64 2.28 9.18 2.52 6.78 4.06 2.07 2.32 3.84 10.80 11.53 7.96 0.74 18.07 20.43 3.87 26.67 5.15 17.09 3.31 2.63 8.24 1.33 0.70 1.31 44.30 14.65 1.18 13.66 6.56 8.47 2.00 1.03 1.34 4.60 11.53 2.14 0.86 0.49

NOM (03) (08) (02) (07) (04) (05) (07) (14) (03) (06) (18) (16) (17) (19) (19) (12) (18) (16) (14) (08) (15) (09) (09) (09) (13) (13) (14) (07) (12) (12) (11) (12) (10) (12) (12) (15) (13) (13) (11) (11) (05) (06) (09) (07) (06) (09) (09) (12) (05) (03) (05) (07) (07) (06) (05) (05) (09) (03) (04) (14)

9.45 4.48 12.25 4.83 1.57 18.93 7.99 2.16 19.57 11.14 2.49 11.77 1.75 1.14 10.07 9.05 4.53 2.91 8.47 7.07 2.56 23.56 32.67 42.25 6.40 23.06 37.96 64.79 36.56 15.24 15.83 42.59 42.38 4.92 11.47 12.55 13.39 32.36 27.29 32.39 2.94 9.88 9.37 6.45 2.58 2.43 17.57 26.21 7.04 58.05 6.50 7.29 4.55 2.50 3.10 23.29 3.43 26.81 4.12 25.77

IDS (07) (09) (06) (12) (08) (09) (05) (06) (10) (10) (16) (11) (07) (14) (09) (11) (06) (04) (18) (05) (24) (22) (21) (19) (14) (16) (21) (21) (24) (18) (20) (16) (20) (20) (20) (18) (17) (21) (19) (18) (07) (08) (11) (08) (07) (07) (15) (15) (08) (08) (09) (05) (05) (08) (06) (11) (05) (12) (06) (16)

0.68 2.72 1.36 1.09 0.65 3.86 1.66 2.30 2.31 1.28 2.21 4.99 2.28 1.08 2.07 12.16 6.94 6.90 22.83 15.52 2.79 4.61 1.52 10.22 1.67 8.12 1.42 1.04 2.89 22.22 29.95 9.47 4.40 2.65 19.38 16.22 18.08 8.51 2.57 7.11 0.77 2.59 2.86 1.36 1.41 0.94 2.01 4.31 1.66 3.12 1.41 3.82 1.84 1.89 1.72 1.29 2.58 1.21 2.04 14.83

(01) (04) (01) (03) (02) (09) (01) (02) (01) (06) (04) (09) (08) (02) (01) (07) (04) (04) (09) (15) (04) (04) (02) (05) (02) (04) (07) (03) (03) (08) (04) (06) (04) (04) (05) (05) (06) (06) (05) (06) (01) (01) (03) (01) (01) (03) (04) (04) (01) (01) (01) (01) (02) (01) (01) (02) (01) (01) (01) (06)

Notes: Figures in parentheses are one-standard errors computed using 300 bootstrap replications of the model. ‘IDS’ in the last column denotes the idiosyncratic shock.

income-absorption effect dominates the exchange rate effect.17 As the income-absorption effect is eroded as GDP declines, the responses of imports eventually become negative. Again, exports and imports for China are at odds with the results for other countries in the region. 17 A positive nominal shock simulates the economy (e.g., an increase in GDP), and the income-absorption effect refers to a resultant increase in imports. The accompanying exchange rate depreciation leads to a decrease in imports, which is referred to as the exchange rate effect.

464

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

GDP

INF

2

4

0

2

-2

0

-4

-2 0

2

4

6

JP TH MY

8

10

TW SG KR

12

14

PH IN HK

16

EXT

4

0

2

CN

4

6

JP TH MY

8

10

TW SG KR

12

14

PH IN HK

16 CN

IMT

4 2

0

0 -4

-2

-8

-4 0

2

4

6

JP TH MY

8

10

TW SG KR

12

14

PH IN HK

16

0

2

CN

4

6

JP TH MY

8

10

TW SG KR

12

14

PH IN HK

16 CN

Fig. 3.1. Responses of the series in levels to world oil price shocks.

Earlier, Figs. 1 and 2 showed that the common factors well accounted for the movements in the variables and that the heterogeneity present across countries was attributable to the asymmetric transmission of common shocks rather than idiosyncratic shocks. The results of an impulse response analysis can shed light on which common shocks mainly cause asymmetric responses of variables. For a better gauge, Fig. 4 depicts the cross-country standard deviations of impulse responses at each forecasting horizon. Among the five common shocks, the regional GDP shock is least responsible for the dispersion of individual GDPs, followed by the world GDP shock. This consolidates our finding that the responses of GDPs are synchronized across countries to regional and world GDP shocks, which are the major determinants of GDP fluctuation. Overall, the evidence supports the feasibility of forming a currency union in the region. The real exchange rate and nominal shocks do not produce much heterogeneity at short horizons, but the effects are

GDP

INF

7

5

5

3

3 1

1 -1

-1 0

2

4

6

JP TH MY

8

10

TW SG KR

12

14

PH IN HK

16

0

2

CN

4

6

JP TH MY

10

TW SG KR

EXT

8

8

12

14

PH IN HK

16 CN

IMT 6

6

2

4

-2

2 0

-6 0

2

4 JP TH MY

6

8 TW SG KR

10

12 PH IN HK

14

16 CN

0

2

4 JP TH MY

Fig. 3.2. Responses of the series in levels to world GDP shocks.

6

8 TW SG KR

10

12 PH IN HK

14

16 CN

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

GDP

INF

4

7

3

5

2

465

1 3

0

1

-1 0

2

4

6

8

JP TH MY

10

TW SG KR

12

14

PH IN HK

16

0

4

6

8

JP TH MY

EXT

7.5

2

CN

12

14

PH IN HK

16 CN

IMT

6

5.0

10

TW SG KR

2

2.5

-2

0.0 -2.5

-6

-5.0

-10 0

2

4

6

JP TH MY

8

10

TW SG KR

12

14

PH IN HK

16

0

2

CN

4

6

8

JP TH MY

10

TW SG KR

12

14

PH IN HK

16 CN

Fig. 3.3. Responses of the series in levels to regional GDP shocks.

significantly amplified as the forecasting horizon increases. They become the two main sources of GDP dispersion at long horizons, while the contribution of these shocks to the forecast error variance of GDP was small. The nominal shock is also largely responsible for the cross-country dispersion in all other variables. The effects are particularly evident in inflation and money growth, which are directly related to monetary policy. Previously, variance decomposition analysis suggested that country-specific idiosyncratic shocks have persistently significant effects on money growth in a number of countries. Taken together, the results indicate that differences in the monetary policies of different countries may be the major source of cross-country heterogeneity in the region. Currently, there is no mechanism or agreement regarding the coordination of monetary policy in East Asia. This stands in contrast to euro countries, which underwent a series of adjustment processes (e.g., ERM and

GDP

REX

10

3 1

6

-1

2

-3 -5

-2 0

2

4

6

JP TH MY

8

10

TW SG KR

12

14

PH IN HK

16

0

2

CN

4

6

JP TH MY

8

EXT

8

10

TW SG KR

12

14

PH IN HK

16 CN

IMT 2

4

0

0

-2

-4

-4 0

2

4 JP TH MY

6

8 TW SG KR

10

12 PH IN HK

14

16 CN

0

2

4 JP TH MY

6

8 TW SG KR

Fig. 3.4. Responses of the series in levels to regional real exchange rate shocks.

10

12 PH IN HK

14

16 CN

466

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

GDP

INF

3

5

1

3

-1 1

-3 -5

-1 0

2

4

6

8

JP TH MY

10

TW SG KR

12

14

PH IN HK

16

0

2

CN

4

6

JP TH MY

EXT

5.0

8

10

TW SG KR

12

14

PH IN HK

16 CN

IMT 4

0.0

0

-5.0

-4

-10.0 -15.0

-8 0

2 JP TH MY

4

6

8 TW SG KR

10

12 PH IN HK

14

16 CN

0

2

4 JP TH MY

6

8 TW SG KR

10

12 PH IN HK

14

16 CN

Fig. 3.5. Responses of the series in levels to regional nominal shocks.

EMS) before the euro came into effect in 1999. Some degree of coordination in monetary policy and national economic policy may be required beforehand to reduce such heterogeneity and to prepare a more favorable environment for the eventual, formal establishment of a currency union in the region. Regarding exports and imports, the magnitude of dispersion is considerable, but this may be partially due to a few peripheral cases exhibiting very large responses, as shown in the impulse response analysis. The notable example is China, where exports and imports show extremely strong responses to most types of shocks. To check this effect, Fig. 4 also reports the results when China is excluded from the export and import blocks. The dispersion of responses is almost halved in both exports and imports. In fact, the impulse response analysis revealed that, save for a few exceptions, exports and imports show the most synchronized responses irrespective of the structural shocks. As exports and imports are important transmission channels of business cycles, close trade linkages between countries have likely contributed to the synchronization of business cycles in the region. 6. Conclusion This paper empirically investigates the co-movements of key macroeconomic variables for 10 major East Asian countries to shed light on the feasibility of a currency union in the region. The working model is a FAVAR that accommodates a large set of 62 variables consisting of six variables for each country, along with world oil prices and world GDP to capture changes in the world economy. Imposing a short-run block recursive structure, two world shocks and three regional shocks are identified to be responsible for driving the co-movements among variables. The paper examines how and to what extent each country responds to these common shocks, focusing on the degree of business cycle synchronization across countries, a key precondition for considering a regional currency union. Empirical results reveal that the common shocks explain most of the variation in the key variables across countries. The contributions of county-specific idiosyncratic shocks are marginal. The great majority of countries in the region also show a qualitatively similar response to the common shocks. Of particular importance is the result that individual GDPs produce synchronized responses to the two main determinants of world and regional GDP shocks. Exports and imports, which are important channels of business cycle transmission, exhibit the most homogeneous responses irrespective of the shocks. Overall, our findings lend support to the synchronization of business cycles across countries, and can be interpreted favorably for the consideration of a currency union in East Asia. Two remaining issues need to be addressed. First, the responses of Chinese exports and imports to virtually all shocks in the model are by far the largest compared to those of other countries. Given the magnitude of Chinese exports and imports, a further examination is warranted in order to assess how such large responses may affect the degree of business cycle synchronization in the region. The second issue is related to the result that nominal shocks in the region produce much less uniform responses. Idiosyncratic shocks also have persistent effects on money growth in a number of countries. Differences in monetary policy and the associated institutions may be the main source of cross-country heterogeneity. On the one hand, it appears that the fluctuations in GDP in response to the world and regional shocks are similar across the 10 countries (a point in favor of a currency union) but on the other hand, the diverse responses of inflation and money growth rates to the regional nominal shock and the findings for the idiosyncratic shock suggest quite

GDP

2.5

DMO

1.2

2.0

1.0

2.00

0.8

1.50

0.5

0.6

1.00

0.0

0.4

1.5 1.0

4

6

8

10

12

WR_OIL

GDP

WR_GDP

REX

14

16

4

NOM

REX

2.5

0.50 2

6

10

12

WR_OIL

GDP

WR_GDP

REX

14

16

2

NOM

EXT

4.5

2.0

8

1.50

2.5

1.25

0.5

1.5

1.00

0.0

0.5

1.0

2

4

6

8

10

12

WR_OIL

GDP

WR_GDP

REX

14

16

4

6

8

10

12

WR_OIL

GDP

WR_GDP

REX

14

16 NOM

IMT

3.5

1.6 2.5

1.2

1.5

0.8 0.4

8

10 GDP

WR_GDP

REX

12

14

16 NOM

IMT W/O China

0.75 2

NOM

INF

6

WR_OIL

1.75

3.5

1.5

4

2

4

6

8

10

WR_OIL

GDP

WR_GDP

REX

12

14

16 NOM

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

2

EXT W/O China

2.50

0.5 2

4

6

8

10

12

WR_OIL

GDP

WR_GDP

REX

14

16 NOM

2

4

6

8

10

12

WR_OIL

GDP

WR_GDP

REX

14

16 NOM

Fig. 4. Dispersion of the responses.

467

468

H. Huh et al. / International Review of Economics and Finance 39 (2015) 449–468

independent monetary policy across the countries (a finding not conducive to a currency union). Such heterogeneities may be smoothed out ex post a monetary union. Yet, some level of coordination in monetary and macroeconomic policy may be necessary as a precursor so as to speed up convergence and prepare a more favorable environment for the introduction of a currency union in East Asia. The evidence presented in this paper, however, does not validate the contention that an East Asian currency union is ready to launch. On a grand scale, there are many more economic and political factors to be considered. As economic conditions for an OCA improve, East Asia may consider further monetary and financial cooperation and integration. Some also argue that OCA criteria are often endogenous (e.g., Frankel & Rose, 1998); that is, joining a monetary union may promote trade integration and capital mobility, thus increasing the degree of symmetry of shocks and business cycle correlations across countries. Moving forward in that direction would require strong political commitments and institutional support, but political cooperation and institutionalization for monetary integration in East Asia are so far seen as relatively weak; this is an important barrier to regional monetary arrangements. Bayoumi et al. (2000) stressed the need for a firm political commitment, and Willet, Permpoon, and Srisorn (2010) added that close attention should be paid to coordination in monetary, fiscal, and exchange rate policies. Indeed, a regional monetary arrangement is only likely to work when there is strong regional solidarity and political support for the delegation of monetary policy to a supra-national institution (e.g., a regional central bank), with systemic support from other regional institutions such as a customs union. Further studies are warranted prior to discussing any formal arrangement of an East Asian currency union to avoid the painful experience of the euro zone. For example, Darvas, Rose, and Szapary (2005) present evidence that business cycles among countries show more synchronized patterns with similar budget positions and with reduced fiscal deficits. Hence, to further test the plausibility of monetary union in East Asia, fiscal variables may need to be considered in the model. Further analysis may be required as suggested by Del Negro and Otrok (2008), to include fiscal variables and/or time-varying analysis, providing larger sample sizes and comparable data frequencies become available. Therefore, we leave them for future research. Acknowledgments The authors thank Thomas Willet, James Morley and Glenn Otto for helpful comments. References Bagliano, F., & Morana, C. (2009). International macroeconomic dynamics: A factor vector autoregressive approach. Economic Modelling, 26, 432–444. Bai, J., & Ng, S. (2002). Determining the number of factors in approximate factor models. Econometrica, 70, 191–221. Bayoumi, T., & Eichengreen, B. (1994). One money or many?: Analyzing the prospects for monetary unification in various parts of the world. Princeton Studies in International Finance76. (pp. 193–229). Princeton University, 193–229. Bayoumi, T., Eichengreen, B., & Mauro, P. (2000). On regional monetary arrangements for ASEAN. Journal of the Japanese and International Economies, 14, 121–148. Bernanke, B., & Boivin, J. (2003). Monetary policy in a data-rich environment. Journal of Monetary Economics, 50, 525–546. Bernanke, B., Boivin, J., & Eliasz, J. (2005). Measuring the effects of monetary policy: A factor-augmented vector autoregressive (FAVAR) approach. Quarterly Journal of Economics, 120, 387–422. Blanchard, O., & Quah, D. (1989). Dynamic effects of aggregate demand and supply disturbances. American Economic Review, 79, 655–673. Boivin, J., & Ng, S. (2006). Are more data always better for factor analysis? Journal of Econometrics, 132, 169–194. Burbidge, J., & Harrison, A. (1984). Testing for the effects of oil-price rises using vector autoregressions. International Economic Review, 25, 459–484. Darvas, Z., Rose, A., & Szapary, G. (2005). Fiscal divergence and business cycle synchronization: Irresponsibility is idiosyncratic. NBER Working Paper No. 11580. Del Negro, M., & Otrok, C. (2008). Dynamic factor models with time-varying parameters: Measuring changes in international business cycles, staff reports. 326, Federal Reserve Bank of New York. Eichenbaum, M., & Evans, C. (1995). Some empirical evidence on the effects of shocks to monetary policy on exchange rates. Quarterly Journal of Economics, 110, 975–1009. Eickmeier, S. (2009). Comovements and heterogeneity in the EURO area analyzed in non-stationary dynamic factor model. Journal of Applied Econometrics, 24, 933–959. Fisher, L., & Huh, H. -S. (2002). Real exchange rates, trade balances, nominal shocks: Evidences for the G-7. Journal of International Money and Finance, 21, 497–518. Forni, M., Hallin, M., Lippi, M., & Reichlin, L. (2000). The generalized factor model: Identification and estimation. The Review of Economics and Statistics, 82, 540–554. Frankel, J., & Rose, A. (1998). The endogeneity of the optimum currency area criteria. The Economic Journal, 108, 1009–1025. He, D., & Liao, W. (2012). Asian business cycle synchronization. Pacific Economic Review, 17, 106–135. Hooker, M. (1996). What happened to the oil price-macroeconomic relationship? Journal of Monetary Economics, 38, 195–213. Kilian, L. (2009). Not all oil price shocks are alike?: Disentangling demand and supply shocks in the crude oil market. American Economic Review, 99, 1053–1069. Kim, D. (2007). An East Asian currency union?: The empirical nature of macroeconomic shocks in East Asia. Journal of Asian Economics, 18, 847–866. Kose, A.M., Otrok, C., & Whiteman, C. (2003). International business cycles: World, region and country specific factors. American Economic Review, 93, 1216–1239. Lee, G.H.Y., & Azali, M. (2012). Is East Asia an optimum currency area? Economic Modelling, 29, 87–95. Lee, G.H.Y., & Koh, S. (2012). The prospects of a monetary union in East Asia. Economic Modelling, 29, 96–102. Lee, J. -H., Park, Y. -C., & Shin, K. -H. (2003). A currency union in East Asia, Institute of Social and Economic Research. Discussion paper no. 571. Osaka University. McKinnon, R. (1963). Optimum currency areas. American Economic Review, 53, 717–725. Moneta, F., & Rüffer, R. (2009). Business cycle synchronization in East Asia. Journal of Asian Economics, 20, 1–12. Morana, C. (2007). Multivariate modelling of long memory processes with common components. Computational Statistics and Data Analysis, 52, 919–934. Mundell, R. (1961). A theory of optimum currency areas. American Economic Review, 51, 657–665. Prasad, E., & Gable, J. (1998). International evidence on the determinants of trade dynamics. IMF Staff Papers, 45, 401–439. Rana, P.B. (2007). Economic integration and synchronization of business cycles in East Asia. Journal of Asian Economics, 18, 711–725. Stock, J., & Watson, M. (2005). Implications of dynamic factor models for VAR analysis. NBER working paper no. 11467. Willet, T., Permpoon, O., & Srisorn, L. (2010). Asian monetary cooperation: Perspectives from the optimum currency area analysis. Singapore Economic Review, 55, 103–124. Yuen, H. (2001). Optimum currency areas in East Asia: A structural VAR approach. ASEAN Economic Bulletin, 18, 206–217.