Exchange Rate Volatility, Exports and Global Value Chains *

Transcription

1 Exchange Rate Volatility, Exports and Global Value Chains * Kiyotaa Sato and Shajuan Zhang February 2017 Abstract Global value chains (GVCs) have become a dominant feature of world trade, and the impact of GVCs on export responsiveness to exchange rate volatility is an important research question. To address how and to what extent GVCs affect a relationship between exchange rate volatility and trade, we conduct a panel estimation using the bilateral and detailed sectoral data covering 29 countries and 18 manufacturing sectors over 1997 to To measure the degree of a country s participation in GVCs, we follow the method developed by Koopman et al. (2014) and Wang et al. (2013) to decompose a bilateral gross trade using the YNU-GIO (Global Input-Output) table. Calculating two measures of GVC participation rates, we investigate whether an effect of exchange rate volatility on exports can be mitigated by the GVC participation. The main finding is that exchange rate volatility has a negative effect on exports, although conditional on the degree of GVC participation. On average, the GVC participation reduces a negative impact of exchange rate volatility on exports by 74 percent. Furthermore, if a country s GVC participation exceeds a threshold level, exchange rate volatility would have a positive effect on the country s exports. JEL Classification: F31, F33, F14 Keywords: Global Value Chains, Global Input-Output Table, Exports, Exchange Rate Volatility * This study is conducted as a part of the Project Exchange Rates and International Currency undertaen at Research Institute of Economy, Trade and Industry (RIETI). The authors would lie to than Eiji Ogawa, Etsuro Shioji, Yushi Yoshida, Willem Thorbece, Craig Parsons, Juno Shimizu, Nagendra Shrestha, Yoshimi Taiyo, and Discussion Paper seminar participants at RIETI. (Corresponding Author) Department of Economics, Yoohama National University, and School of Business and Law, Edith Cowan University. sato@ynu.ac.jp Center for Economic Growth Center, Yoohama National University. zhangshajuan@gmail.com 1

2 1. Introduction A traditional criticism of flexible exchange rate regimes is that exchange rate volatility tends to reduce international trade by inducing uncertainty into international transactions. However, earlier studies showed that a negative impact of exchange rate volatility on trade is theoretically and empirically ambiguous (Auboin and Ruta, 2012). Recent studies have emphasized that the ambiguous results may be due to an aggregation bias (Byrne et al., 2008), and found a more robust negative relationship between exchange rate volatility and trade flows using a disaggregated data (Thorbece, 2008; Hayaawa and Kimura, 2009; Tang, 2014; and Sato et al., 2016). As discussed in Auboin and Ruta (2012), the relationship between the exchange rate volatility and trade is found to be affected by various factors, such as the existence of hedging instruments, production structures, the degree of economic integration across countries, etc. In addition, Global Value Chains (GVCs) that are vigorously discussed in recent years need to be considered as well when analyzing the relationship between the exchange rate volatility and trade, because of growing cross-country transactions of intermediate goods and final goods exports along GVCs. GVCs have become a dominant feature of world trade (Hummels et al., 2001; Feenstra and Jensen, 2009). Figure 1 shows the development of GVCs for different regions. The measurement of a country s participation in GVCs by region is constructed in this study by following Koopman et al. (2010). 1 Although a temporary fall in GVCs is observed during the period of the global financial crisis in , GVCs exhibit a steady growth over the sample period. Given increasing importance of GVCs, this paper addresses a question as to how and to what extent GVCs affect the export responsiveness to the exchange rate volatility. Figure 2 shows a relationship between nominal exchange rate volatility and exports in 2012 for countries at a different level of GVC participation in manufacturing sectors. 2 Countries are raned based on their level of GVC participation measured by the method developed by Koopman et al. (2010). The left-hand side shows the lower quartile, while the right-hand side shows the upper quartile of the distribution. Each case provides a comparison of the residuals of a real export regression on a set of variables and the residuals of an exchange rate volatility regression on the same variables using the data as of Control variables include real GDP of importers and 1 The details of GVC participation measurements are discussed in Section 2. We use PGVC1 presented in Section 2 for Figure 1. 2 Again, we use PGVC1 for Figure 2. The details of this GVC measurement are presented in Section 2. 2

3 exporters, real bilateral exchange rates. Year effect and importer-exporter-sector effect are also included. There is a slightly negative relationship between exports and the exchange rate volatility for countries that have relatively low participation in GVCs, whereas there is a slightly positive relationship for countries that are more deeply involved in GVCs. This result shows a preliminary evidence that the effect of exchange rate volatility on exports varies with the degree of GVC participation, which motivates our research. In our empirical analysis, we use a bilateral and sectoral data to avoid aggregation bias, and employ a panel framewor covering 29 countries and 18 manufacturing sectors over the period from 1997 to The sample period and countries are determined by the availability of trade data. We use YUN-GIO (Global Input-Output) table to measure the GVC participation rate, because YNU-GIO table covers eleven Asian countries as an endogenous country, 3 and hence much better than World Input-Output Database (WIOD) that includes just six Asian countries when analyzing a dynamic regional integration process in Asia. As shown in Figure 1, an average of the GVC participation rate is relatively higher in Asia than others, indicating that GVCs are well established in Asian region. Given the importance of cross-border production sharing in manufacturing sectors, we focus on manufacturing exports. As for the exchange rate volatility, we use the nominal exchange rate volatility for two reasons. First, the nominal exchange rate is a policy instrument that policy maers can directly affect, and hence more relevant to policy implications. Second, sectoral price data is necessary to calculate sectoral real exchange rates, but is not available for all sample countries. If we used aggregated price data, such as consumer price index (CPI), for a sectoral level analysis, we would fail to avoid aggregation bias (Byrne et al., 2008) and to capture a difference in sectoral price trends when calculating bilateral real exchange rates. How do GVCs affect the effect of exchange rate volatility on trade? GVCs are expected to lower the responsiveness of exports to exchange rate volatility for two reasons. First, exchange rate changes tend to have an offset effect. Specifically, exchange rate depreciation can improve export competitiveness of domestic value added on one hand, but raise the cost of imported inputs on the other hand Second, a partner 3 YNU-GIO table is published by Center for Economic and Social Studies (CESSA), Yoohama National University ( Shrestha and Sato (2015) investigates the degree of global and regional transmissions in export demand shoc by using three databases of globally lined input-output tables: YNU-GIO, WIOD, and OECD Inter-Country Input-Output (ICIO) tables. Shrestha and Sato (2015) demonstrated that a difference in the degree of shoc transmission between three databases rests on a coverage of endogenous countries. 3

4 relationship in production networs is quite stable because of the relation-specific nature of intermediate goods transactions. Once production networ is established, it is not easy for firms to switch foreign suppliers or to find new buyers in response to exchange rate changes (Obashi, 2010). However, there is another argument that GVCs augment the negative effect of exchange rate volatility on trade, because trade costs caused by exchange rate uncertainty are liely to be more relevant in supply chain trade. Hayaawa and Kimura (2009) and Tang (2014) found that a negative effect of the exchange rate volatility on exports is more liely to occur in intermediate goods trade. Whether GVCs have a positive or negative effect on the relationship between exchange rate volatility and trade depends on which effect is dominant. This is still an open question, and more empirical wor needs to be conducted. One of novelties of this paper is to employ two measures of participation in GVCs. A country s participation in GVCs can tae place through forward-linage or/and bacward-linage. GVC participation through forward-linage means that a country s domestic value added embodied in intermediate goods is re-exported to a third country. Participating in GVCs through bacward-linage means that a country uses imported inputs to produce export goods. To trace such transactions of intermediate goods, we first decompose bilateral sectoral gross exports not only by employing the YNU-GIO table, but also by following the method developed by Koopman et al. (2014) and Wang et al. (2013). Figure 3 illustrates the basic data of gross export decomposition. Then, we augment the method developed by Koopman et al. (2010) to build the first measure of bilateral participation in GVCs, that is, a sum of forward-linage and bacward-linage. The second measure is constructed by using the method proposed by Wang et al. (2013), which focuses on the pure double counting terms. The higher the value of these two measures, the deeper the GVC participation is. But, as discussed above, its impact of GVC participation on a relationship between exchange rate volatility and trade is basically ambiguous. We employ a dynamic panel model that can capture the dynamic nature of trade. 4 To address the issues of joint endogeneity of all explanatory variables, we use the two-step system GMM estimator developed by Arellano and Bond (1991), Arellano and Bover (1995), and Blundell and Bond (1998) with robust standard errors computed by the methodology proposed by Windmeijer (2004). To investigate the effect of GVCs on the relationship between exchange rate volatility and exports, we add an interaction term to the basic regression that is widely used in the literature in order to allow the 4 A firm s decision to export depends on its prior experience in the export-maret (Roberts and Tybout,1997; Das et al., 2007). 4

5 trade effect of exchange rate volatility to vary across countries and sectors. Our paper is related to two strands of research. First, it relates to the empirical literature on the impact of exchange rate volatility on exports. This study empirically investigates the role of GVCs in a relationship between exchange rate volatility and exports by using two measures of bilateral and sectoral production sharing. We contribute to the literature by showing that the extent to which countries participate in GVCs is also an important factor in explaining the effect of the exchange rate volatility on exports. Second, our paper is related to the recent studies that consider a relationship between exchange rate changes and GVCs, such as Eichengreen and Tong (2015), Ahmed et al. (2015) and Cheng et al. (2016). Eichengreen and Tong (2015) found that Renminbi appreciations have a positive effect on the valuations of firms exporting final goods to China, but little effect on those exporting intermediate goods for China's processing exports. Ahmed et al. (2015) empirically examined how the formation of GVCs affects the exchange rate elasticity of exports and found that GVCs participation reduces the real effective exchange rate (REER) elasticity of manufacturing exports by 22 percent on average. Cheng et al. (2016) found the size of exchange rate elasticity of exports is smaller when the import content of GVC exports is larger. Different from these studies, we analyze the effect not of the exchange rate elasticity but of the exchange rate volatility on exports, and demonstrate that such effect is significantly affected by the degree of a country s participation in GVCs. 5 There are three main findings in this paper. First, exchange rate volatility has a negative effect on exports, but this effect is very small. Second, GVC participation reduces the negative impact of exchange rate volatility on exports. That is, the impact of exchange rate volatility on manufacturing exports will be reduced by 74 percent on average. Third, if the degree of a country s participation in GVCs exceeds the threshold level, exchange rate volatility would have a positive effect on that country s exports. The remaining part of this paper is organized as follows. Section 2 discusses the research methodology, definition of variables and description of the data. Section 3 presents our estimated results, and robustness chec is conducted in Section 4. Finally, Section 5 concludes this study. 2. Empirical methodology 5 A theoretical analysis of the effect of GVCs on a lin between exchange rate volatility and exports is not made in this paper but will be conducted for future research. 5

6 2.1 Model Specification Given the recent growth of GVCs in world trade, an interesting question is whether a level of engagement in GVCs conditions the impact of exchange rate volatility on exports. This is still an open question. To investigate the role of GVC participation in the responsiveness of exports to exchange rate volatility, we consider the standard regression that is widely used in the literature. We add to the regression a measure of exchange rate volatility, its interaction term with GVC participation, and other controls. To capture the dynamic nature of trade, we employ a dynamic panel model and use the two-step system GMM estimator developed in Arellano and Bond (1991), Arellano and Bover (1995), and Blundell and Bond (1998) with robust standard errors computed by the methodology proposed by Windmeijer (2004). One advantage of this approach is to address the issues of joint endogeneity of all explanatory variables. We start our analysis by examining the direct effect of exchange rate volatility on real exports, then proceed to loo at interactions between these measures and GVC participation. More specifically, the following equation is estimated: ln Exp 0 1 ln Exp 1 2Vol 3Vol * PGVC 4 1 ' D u ij t PGVC 1 (1) where, i, j and t denote sector, exporter, importer and time, respectively; ln is a natural logarithm of real exports, which is deflated by CPI. is the bilateral nominal exchange rate volatility between country i and country j. is a country s participation in GVCs in year t 1. We use the lagged variable to account for a possibility that PGVC may be endogenous to the exchange rate volatility. To ensure that PGVC is exogenous to the exchange rate volatility, we use two alternatives. First is an average of PGVC over time, symbolled as PGVC_mean. Second is a value of PGVC in the first sample year, symbolled as PGVC_1997. is a set of other control variables commonly used in the literature, including real GDP of exporters and importers, bilateral real exchange rate defined as the price of a country j s currency in terms of a country i's currency to capture the relative price effect, and an interaction term of the bilateral real exchange rate with GVC participation. is an unobservable exporter-importer-sector specific effect which captures time invariant and country-specific, industry-specific and country-pair factors, so that we do not include time invariant gravity variables, such as a distance, an adjacency dummy, and a 6

7 language dummy into this specification of the model; is a time-fixed effect which controls for all common shocs to all country-pairs and industries, such as changes in world demand, technological change, oil price shocs. is an error term. Our hypothesis is that <0 and <or>0. Participation in GVCs could affect the relationship between exchange rate volatility and exports in both two directions. One argument is that participation in GVCs will reduce a negative effect of exchange rate volatility on trade due to the offsetting effect of exchange rate changes and the stability of production networ. Another argument is that a participation in GVCs will magnify the negative effect of exchange rate volatility on exports, because trade costs, due to exchange rate uncertainty, may be more relevant in supply chain trade. Therefore a sign of depends on which effect is more dominant. If <0, the impact of exchange rate volatility,, is more negative at a higher level of participation in GVCs. If >0, a higher level of participation in GVCs will reduce the negative impact of exchange rate volatility on real exports. Moreover, when and have an opposite sign, a threshold value will be: ln Exp Vol 2 3PGVC 0 PGVC 2 3 (2) 2.2 Data and Variables A panel dataset covers 29 countries and 16 manufacturing sectors over the period from 1997 to The sample period, sample countries, and sector coverage are determined by the availability of trade data. All trade data for assessing the role of participation in GVCs are taen from the YNU-GIO table. A major advantage of this dataset is that it covers 11 Asian countries as an endogenous country, while WIOD, which is widely used in the literature, covers only five Asian countries. Since the GVCs are well established in Asian region, we need to include as many Asian countries as possible in the empirical analysis. For robustness chec, we also use the trade data obtained from WIOD to re-estimate equation (1). Given the importance of cross-border linages in manufacturing sector, we focus on exports in 16 manufacturing sectors. The details of sector classifications are shown in Table 1. The GDP data is taen from World Ban, World Development Indicators (WDI). All other data are obtained from IMF, International Financial Statistics. We use a moving sample standard deviation of the log-differenced bilateral 7

8 nominal exchange rate as a measure of exchange rate volatility, which is widely used in the literature. As discussed in Clar et al. (2004) and Thorbece (2008), when considering the impact of exchange rate volatility on trade flows, the timing issue is crucial. Clar et al. (2004) states that firm s trade is less responsive to the short-term volatility of exchange rates, given a large investment of exporting firms in foreign marets to build mareting and distribution networs and/or to set up production facilities. This paper uses five inds of time windows to allow for timing and uncertainty issues. The first one is the exchange rate volatility in the current year, which is called short-term volatility (i.e., one-year volatility). The second one is the volatility during the current and the previous year (i.e., two-year volatility), and the third one is the volatility in the current year and the previous two years (i.e., three-year volatility). The second and third indicators are for the medium-term volatility. The fourth one is the volatility during the current and previous four year (i.e., five-year volatility), which is for the long-term volatility. The fifth one is the volatility in the previous year, the current year and the next year. Thorbece (2008) uses this fifth measure to allow for uncertainty of exchange rate changes, since uncertainty is a forward-looing concept (i.e., two year and forward volatility). We use two measures of GVC integration to assess the GVC effects on the relationship between the exchange rate volatility and trade. To build these measures, we first decompose the bilateral sectoral gross exports into 16 terms as shown in Figure 3. This approach is developed by Koopman et al. (2014) and Wang et al. (2013). The details of each term are shown in Table 2. Then, we augment the method developed by Koopman et al. (2010) to build the first bilateral participation index at a sectoral level when using this decomposition. To build this measure, we define the forward linage (_ ) and the bacward linage (_ ) as follows: PGVC PGVC _ FL _ BL FL Expt BL Expt FL jit Expt BL jit Expt jit jit where ( ) is the value of inputs produced in country i (j) embodied in sector s gross exports to country j (i) that used in country j (i) s exports; 6 ( ) is the foreign content in country i (j) s gross exports to country j (i) in sector ; and and correspond to the total amounts of T3, T4, and T5 in Table 2. correspond to the total amounts of T11 through T16 in Table 2. The bacward 8

9 ( ) is country i (j) s gross exports to country j (i) in sector. One country can engage in GVCs either through forward-linage or bacward-linage. Thus, we define the first bilateral participation index as sum of forward-linage and bacward-linage: PGVC ( 1) FL BL Expt FL jit BL Expt jit jit The second measure of participation in GVCs is computed by using the method proposed by Wang et al. (2013). The formula is given by: PGVC ( 2) FDC BL FDC BL jit jit where ( ) is a pure double counting of foreign value added (FDC, including country j (i) value added) in country i (j) s gross exports to country j (i) at sector (i.e., total amount of T15 and T16 in Table 1). In GVCs, intermediate goods are traded across national borders multiple times before they are used to produce final goods production. FDC can happen when intermediate goods are traded bac and forth between countries. The more bac and forth trade, the more deeply the countries are involved in GVCs, in which case FDC becomes larger. Thus, an increasing weight of FDC in BL indicates deepening cross-country production sharing. Summary statistics of main variables are presented in Table 3. For further evaluation of the role of GVC integration, we also report 10 percentile, 50 percentile and 90 percentile of each variable. A scale of two measures (PGVC1 and PGVC2) of participation in GVCs is quite different. An average of PGVC1 is 0.353, and an average of PGVC2 is A mean value of bacward linage is and somewhat higher than that of forward linage (0.151). 3. Empirical Results In order to test whether and to what extent a GVC participation affects the impact of exchange rate volatility on exports, we use the two measures of GVC participation. We include each measure in equation (1) in three different ways: (i) a linage is also labeled as Vertical Specialization (VS) in Hummels et al. (2001). 9

10 lagged GVC measure, (ii) a GVC measure of the first sample year, and (iii) an average GVC measure over the whole sample period, to capture a possibility that GVC integration may be endogenous to the exchange rate volatility. We also use the five measures of exchange rate volatility, and the result based on the five-year volatility of exchange rate is treated as a benchmar result. Tables 4 and 5 present estimated results of equation (1) by using the two different measures of GVC participation. Each table displays the results of seven regressions. The first regression estimates the effect of exchange rate volatility without interaction terms. The second to the fourth regressions add a variable interacting between the exchange rate volatility and the measure of GVC participation. We add the interaction term of exchange rate volatility with the lagged measure of GVC participation in the second regression. The interaction term of the exchange rate volatility and the measure of GVC participation in the first sample year are included in the third regression. The interaction term of exchange rate volatility and the averaged GVC participation is included in the fourth regression. We also add the level of real exchange rate and its interaction term with GVCs, and the results are reported in columns (5) through (7). Column (1) of Table 4 illustrates the negative effect of exchange rate volatility on real exports but its effect is quite small, which is consistent with the findings of previous studies (See survey in Auboin and Ruta, 2012). Columns (2) through (4) show that the interaction term of exchange rate volatility with GVC participation is positive and significant at 1% level, which indicates that the more deeply a country is engaged in GVCs, the lower the point estimate of the exchange rate volatility effect is on exports. This result is consistent with Sato et al. (2016), which shows that effect of exchange rate volatility on trade differs across industries and that the exchange rate volatility has a weaer impact on the industries where GVCs are relatively well established. The results in columns (5) through (7) show that the real exchange rate and its interaction term with the GVC participation do not exhibit any significant coefficients, while the effects of exchange rate volatility and its interaction term with the GVC participation do not change. The above results have important implications for the effect of exchange rate on exports. Traditional macroeconomic models tend to predict that a currency depreciation promotes exports through expenditure switching mechanism by changing the relative price between export and local products. However, we have found that real exchange rate does not have any significant impact on exports. In contrast, we have revealed that the exchange rate volatility has significantly negative effect on exports. 10

11 Since the exchange rate volatility itself taes an opposite sign of coefficient to its interaction term with the GVC participation, we can compute a threshold level of the GVC participation, above which the exchange rate volatility promotes exports. In the bottom part of Table 4, we present the estimated threshold levels ranging from to 0.654, which are above the mean value of the GVCs measures reported in Table 3. In Table 5, we use the second measure of GVC participation proposed by Wang et al. (2013). The results of estimation are almost the same as those of Table 4, but there are a few differences to be discussed. First, the coefficient of GVC participation itself is significantly positive only at the 10% level in column (1), and the coefficient of interaction term between the exchange rate volatility and GVC participation is significantly positive only at the 10% level if including the real exchange rate and its interaction term in column (4). However, in columns (1) and (4), we use the level of GVC participation in the previous year, which may be less reliable than the other two GVC participation variables. The level of GVC participation in 1997 and the average level of GVC participation are less liely to be affected by the exchange rate volatility, and hence they are more exogenous to the exchange rate volatility. All single terms of exchange rate volatility are significantly negative even at 1% level and their magnitude remains almost unchanged. Second, all interaction terms of exchange rate volatility and participation in GVCs are significantly positive at least at 5% level except for column (4), and the magnitude of those is quite similar to those from table 4. These results suggest that the positive impact of GVC participation on the trade effect of exchange rate volatility remains robust with different measures of GVC participation. To mae further investigation, we proceed to examine the impact of bacward-linage and forward-linage on the trade effect of exchange rate volatility. The results are shown in Table 6. While columns (1) and (4) show that the bacward-linage does not have any impact on the relationship between exchange rate volatility and exports, but the forward-linage has significantly positive effect when lagged measures of bacward-linage and forward-linage are included in equation (1). Turning to results from the other two alternatives (the GVC measure of the first sample year and the average GVC measure over the whole sample period), which are considered more exogenous to the exchange rate volatility, both bacward-linage and forward-linage have significantly positive effect. Thus, we can conclude that a country s GVC participation either through bacward linage or forward linage can reduce the negative effect of exchange rate volatility on exports. To investigate the role of GVC integration, we allow the trade effect of exchange rate volatility to vary with the level of a country s participation in GVCs by 11

12 adding interaction terms to the model captured by. We can evaluate the role of GVC participation using a formula, /, with a specific value of PGVC. Since we have lots of estimated values of and, we tae an average of those estimates in each table. For example, Table 4 shows that the average point estimates of the impact of exchange rate volatility for countries with no GVC participation ( ) is , 8 and the average point estimates of interaction terms ( ) is Therefore, on average, a country s GVC participation (PGVC1) can reduce the negative trade effect of exchange rate volatility by approximately 74 percent (0.046*0.353/0.022). We can do the same calculation for 10 percentile, 50 percentile, and 90 percentile value of each measure of the GVC participation using estimated values from Tables 4 to 6. The results are presented in Figure 4. GVC participations reduce the effects of exchange rate volatility on exports by approximately 58 to 69 percent for those countries whose degree of GVC participation is around at a median level. This effect is economically important, indicating that GVC participation plays a large role in the relationship between exchange rate volatility and exports. It is very interesting that Figure 4 shows that with 90th percentile of participation in GVCs( PGVC1 and PGVC2), a reduction effect of each measure of GVC participation becomes 125% and 123%, respectively. This means that if a country is involved in GVCs at 90th percentile level, exchange rate volatility will have a positive effect on exports. We also do the same calculation for the 10 th, 50 th and 90 th percentile of bacward- and forward-linages using the estimates in Table 5. Figure 4 shows that bacward linage reduces the negative trade effect of exchange rate volatility by 47 percent and forward linage reduces the effect by 27 percent at a median level of each measure. We can also quantify the impact of exchange rate volatility. Figure 5 illustrates when the exchange rate increases by one standard deviation (i.e., 0.192), 10 how exports will change with no participation in GVCs, 10 percentile of participation in GVCs, 50 percentile of participation in GVCs, and 90 percentile of participation in GVCs, respectively. Figure 5 shows that if a country does not participate in GVCs, we would expect that one standard deviation of exchange rate volatility would reduce real exports by approximately 0.30%. 11 This effect is economically small and can be reduced by deepening cross country production sharing. In the case of 90 percentile of participation in GVCs, one standard deviation of exchange rate volatility would increase the 8 -( )/6= ( )/6= is taen from the standard deviation of the five-year exchange rate volatility reported in Table The average of estimates based on two measures of participation: (0.41%+0.22%)/2 12

13 country s real exports by 0.08% (average effect of PGVC1 and PGVC2). 4. Robustness Chec To chec the robustness of our main result, we conduct additional empirical examinations. First, we re-estimate equation (1) using four different time windows of exchange rate volatiltiy: (i) the exchange rate volatility in the current year (i.e., one-year volatility), (ii) the exchange rate volatility during the current and the previous year (i.e., two-year volatility), (iii) the exchange rate volatility in the current year and the previous two years (i.e., three-year volatility), and (iv) the exchange rate volatility in the previous year, the current year and the next year (i.e., two year and forward volatility),. The results are presented in Table 7. Second, to rule out any disturbances caused by the currency crisis in 1997 and the global financial crisis in 2008, we re-estimate equations with a sub-sample from 1999 to 2007, and the results are presented in Table 8. Third, we re-estimate the model by using trade and IO data taen from WIOD, and the results are presented in Table 9. These tables show estimated coefficients for the main variables: the real exchange rate volatility and its interaction term with different measures of GVC participation. Table 7 shows the results with different time windows of exchange rate volatility. Column (1) shows that the interaction terms of the volatility with GVC participation are insignificant in most regressions when the exchange rate volatility is measured using the current year only, which indicates that the GVC participation does not affect the effect of exchange rate volatility on exports in the short-run. Turing to the measures of longer time windows, almost all estimates of interaction terms are significant and their magnitudes become larger for longer time windows. Thus, the GVC participation will have positive impact on exports in the medium- and long-run. Table 8 presents the results using the sample period from 1999 to The estimated results are very similar to our benchmar result, which suggests that our results are quite robust. Even we control for the disturbances caused by the currency crisis in 1997 and the global financial crisis in 2008, the GVC participation remains to have a significantly positive effect on the relationship between exchange rate volatility and exports. Table 9 presents the results using the trade data taen from WIOD to chec whether we can obtain the similar results even when employing different database from YNU-GIO table. The estimated results are quite similar to the results of Tables 4 and 5, 13

14 except for the case when estimating the interaction term of exchange rate volatility and the lagged GVC participation. The estimated interaction term is significantly negative as shown in Table 9. As we discussed earlier, however, the GVC participation may be endogenous to the exchange rate volatility, while the two alternative measures, the GVC measure of the first sample year and the average GVC measure over the whole sample period, are considered more exogenous to the exchange rate. Thus, our main results can be supported even when using different data sources. 5. Concluding Remars In this paper, we have empirically investigated how GVC participation affects the impact of exchange rate volatility on exports using the YNU-GIO table that includes more endogenous Asian countries than WIOD. We employed the dynamic panel model by using the two-step system GMM estimator developed by Arellano and Bond (1991), Arellano and Bover (1995), and Blundell and Bond (1998) to deal with possible endogeneity problem. We built two measures of the bilateral participation in GVCs: one is to measure the extent of both forward- and bacward-linages, developed by Koopman et al. (2010). The other is to capture the pure double counting of foreign value added caused by bac and forth trade of intermediate goods between two countries, developed by Wang et al. (2013). The latter measure is quite new and this is the first study that uses the measure for an empirical analysis. There are three main findings in this paper. First, exchange rate volatility has a negative effect on exports, but this effect is very small. Second, GVC participation reduces negative impact of exchange rate volatility on exports. GVC participation reduces the impact of exchange rate volatility on manufacturing exports by 74 percent on average. Third, if a country s participation in GVCs above the threshold level, exchange rate volatility would have a positive effect on that country s exports. The policy implications of our results are, first, countries with no participation to GVCs or less degree of participation in GVCs may benefits from maintaining a rigid exchange rate regime; second, for those countries deeply involved in GVCs (above threshold level), there is no fear of floating exchange rate regime; third, a country deeply involved in GVCs cannot use exchange rate as instrument to boost exports due to the offset effect of changes in exchange rate. 14

15 References Ahmed, S., M. Appendino and M. Ruta (2015), Global Value Chains and the Exchange Rate Elasticity of Exports. IMF Woring Paper. Arellano, M. and S. Bond (1991), Some tests of specification for panel data:monte Carlo evidence and an application to employment equations. Review of Economic Studies, 58, Arellano, M. and O. Bover (1995), An other loo at the instrumental-variable estimation of error-components models. Journal of Econometrics, 68, Auboin, M. and M. Ruta (2012), The Relationship between Exchange Rates and International Trade: A Review of Economic Literature. CESIFO Woring Paper, No Blundell, R. and S. Bond (1998), Initial Conditions and Moment Conditions in Dynamic Panel Data Models. Journal of Econometrics, 87(1), Byrne, J. P, J. Darby and R. MacDonald (2008), US Trade and Exchange Rate Volatility: A Real Sectoral Bilateral Analysis. Journal of Macroeconomics, 30(1), Cheng, K. C., G. H. Hong, D. Seneviratne and R. v. Elan(2016), Rethining the Exchange Rate Impact on Trade in a World with Global Value Chains. International Economic Journal, 30(2), Clar, P. B., N. Tamirisa, S. J. Wei, A. Sadiov and L. Zeng (2004), A New Loo at Exchange Rate Volatility and Trade Flows. IMF Occasional Paper 235, International Monetary Fund, Washington DC. Das, S., M. J. Roberts and J. R. Tybout (2007), Maret Entry Costs, Producer Heterogeneity, and Export Dynamics. Econometrica, 75(3), Eichengreen, B. and H. Tong (2015), Effects of Renminbi Appreciation on Foreign Firms: The Role of Processing Exports. Journal of Development Economics, 116, Feenstra, R. and J. B. Jensen (2009), Evaluating Estimates of Materials Offshoring from U.S. Manufacturing. NBER Woring Paper Hayaawa, K. and F. Kimura (2009), The effect of exchange rate volatility on international trade in East Asia. Journal of The Japanese and International Economies, 23(4), Hummels, D., J. Ishii and K.M. Yi (2001), The Nature and Growth of Vertical Specialization in World Trade. Journal of International Economics, 54(1),

16 Koopman, R., W. Powers, Z. Wang, S.J. Wei (2010), Give Credit Where Credit Is Due: Tracing Value Added in Global Production Chains. NBER Woring Paper No Koopman, R., Z. Wang, and S.-J. Wei (2014), Tracing Value-Added and Double Counting in Gross Exports. American Economic Review, 104(2), Obashi, A. (2010), Stability of production networs in East Asia: duration and survival of trade. Japan and the World Economy, 22(1), Roberts, M. J. and J. R. Tybout (1997), The Decision to Export in Colombia: An Empirical Model of Entry with Sun cost. The American Economic Review, 87(4), Sato, K., J. Shimizu, N. Shrestha and S. Zhang (2016), Industry-specific Exchange Rate Volatility and Intermediate Goods Trade in Asia. Scottish Journal of Political Economy, 63(1), Sato, K., and N. Shrestha (2014), Global and Regional Shoc Transmission: An Asian Perspective. CESSA Woring Paper, Shrestha, N. and K. Sato (2015), Global Financial Crisis and Shoc Transmission: Supply- or Demand-driven Shoc? CESSA Woring Paper, Tang, H. C. (2014), Intra-Asia Exchange Rate Volatility and Intra-Asia Trade: Evidence by Type of Goods. The World Economy, 37(2), Thorbece W. (2008), The effect of exchange rate volatility on fragmentation in East Asia: Evidence from the electronics industry. Journal of The Japanese and International Economies, 22(4), Wang, Z., S.J. Wei and K. Zhu (2013), Quantifying International Production Sharing at the Bilateral and Sector Level, NBER woring paper. Windmeijer, F. (2004), A finite sample correction for the variance of linear two-step GMM estimator. Journal of Econometrics, 126(1),

17 Figure 1 Emergence of Global Value Chains on Manufacturing Sector 0.55 Manufacturing World Asia EU Other Note: Authors calculation. The measure of a country s participation in GVCs is following Koopman et al. (2010). Then, we tae an average of all countries participation for each region. Figure 2 Exchange Rate Volatility and Exports in 2012 Lower Quartile of Integration to GVCs Upper Quartile of Integration to GVCs Real Exports Real Exports Exchange Rate Volatility Exchange Rate Volatility Note: For each quartile, the residuals of a real exports regression on a set of variables are regressed on the residuals of an exchange rate volatility regression on the same variables using data in Control variables are real GDP of importer and exporter, real bilateral exchange rate. Year effect and importer-exporter-sector effect are also included. 17

18 Figure 3 Gross Exports Decompositionn Source: Wang et al. (2013). Figure 4 Impact of GVCss on Trade Effect of Exchange Rate Volatility 140% 120% 125% 123% 100% 80% 94% 10th Percentile 60% 40% 69% 39% 58% 47% 55% 50th Percentile 90th Percentile 20% 24% 26% 27% 15% 0% Participation PGVC1 Participation PGVC2 Bacward Forward 18

19 Figure 5 Trade Effect of Exchange Rate Volatility 0.20% 0.10% 0.00% -0.10% -0.20% -0.30% No GVC Participation 10th Percentile 50th Percentile 90th Percentile -0.40% -0.50% Participation PGVC1 Participation PGVC2 Bacward Forward Table 1 Industry Classification YNU-GIO ISIC.rev3 Industry Name Description Y Food Food, Beverage, Tobabcoo Y Textile Textiles, Textile Products, Leather and Footwear Y05 20 Wood Wood Products(excl. furniture) Y Paper Paper, Paper Products, Printing and Publishing Y07 23 Petroleum Coe, Refined Petroleum Products,Nuclear Fuel Y08 24 Chemical Chemicals and Chemical Products Y09 25 Rubber Rubber and Plastics Products Y10 26 Non-Metal Non-metallic Mineral Products Y11 27 Metals Basic Metals Y12 28 Fabricated M. Fabricated Metal Products Y13 29 General M. Machinery and Equipment n.e.c. Y14 30 Office M. Office,Accounting and Computing Machinery Y15 31 Electrical M. Electrical Machinery and Apparatus n.e.c. Y16 32 Communication Communication Equipment and Apparatus Y17 33 Optical I. Medical, Precison and Optical Instruments Y18 34 Motor Motor Vehicles Y19 35 Transport E. Other Transport Equipment Y Others Other Manufacturing Source: Author s edition based on Sato and Shrestha (2014). 19

20 Table 2 Definition of Decomposition terms Source: Wang et al. (2013). Table 3 Summary of Main Variables Variable Obs Mean Std. Dev P10 P50 P90 Log of Real Exports Participation in GVCs (PGVC1) Participation in GVCs (PGVC2) Bacward Linage Forward Linage Exchange Rate Volatility (One-Year) Exchange Rate Volatility (Two-Year) Exchange Rate Volatility (Three-Year) (Five-Year) Exchange Rate Volatility (Two Year and Forward) Exchange Rate Volatility

21 Table 4 Trade Effect of Exchange Rate Volatility Conditional on GVC Participation Variables (1) (2) (3) (4) (5) (6) (7) Exchange Rate Volatility ** *** *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) ( ) Exchange Rate Volatility Participation in GVCs(PGVC1_t-1 ) Exchange Rate Volatility Participation in GVCs(PGVC1_1997) Exchange Rate Volatility Participation in GVCs(PGVC1_mean) Real Exchange Rate Participation in GVCs(PGVC1_t-1 ) Real Exchange Rate Participation in GVCs(PGVC1_1997) Real Exchange Rate Participation in GVCs(PGVC1_mean) *** *** ( ) ( ) *** *** ( ) ( ) *** *** ( ) ( ) ( ) ( ) ( ) Real Exchange Rate ( ) ( ) ( ) Participation in GVCs (PGVC1_t-1 ) *** 0.133*** ( ) ( ) Lag of Real Exports x x x x x x x Real GDP of Importer x x x x x x x Real GDP of Exporter x x x x x x x Year Effects x x x x x x x Exporter-Importer-Sector Effects x x x x x x x Serial correlation First-order Second-order Threshold Value of GVCs ( ) No.observations Note: Robust standard errors in parentheses, * 10% significance level; ** 5% significance level; ***1% significance level. 21

22 Table 5 Trade Effect of Exchange Rate Volatility Conditional on Alternative Measure of GVC Participation Manufacturing Exports Variables (1) (2) (3) (4) (5) (6) Exchange Rate Volatility *** *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) Exchange Rate Volatility Participation in GVCs(PGVC2_t-1 ) Exchange Rate Volatility Participation in GVCs(PGVC2_1997) Exchange Rate Volatility Participation in GVCs(PGVC2_mean) Real Exchange Rate Participation in GVCs(PGVC2_t-1 ) Real Exchange Rate Participation in GVCs(PGVC2_1997) Real Exchange Rate Participation in GVCs(PGVC2_mean) ** * (0.0114) (0.0112) *** *** ( ) ( ) *** *** (0.0105) (0.0103) ** ( ) * ( ) * ( ) Real Exchange Rate ( ) ( ) ( ) Participation in GVCs (PGVC2_t-1 ) * ** (0.0150) (0.0139) Lag of Real Exports x x x x x x Real GDP of Importer x x x x x x Real GDP of Exporter x x x x x x Year Effects x x x x x x Exporter-Importer-Sector Effects x x x x x x Serial correlation First-order Second-order Threshold Value of GVCs ( / ) No.observations Note: Robust standard errors in parentheses, * 10% significance level; ** 5% significance level; ***1% significance level. 22

23 Table 6 Trade Effect of Exchange Rate Volatility Conditional on Bacward-linage and Forward-linage Manufacturing Exports Variables (1) (2) (3) (4) (5) (6) Exchange Rate Volatility *** *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) ( ) Exchange Rate Volatility Bacward Participation in GVCs_t-1 Exchange Rate Volatility Forward Participation in GVCs_t-1 Exchange Rate Volatility Bacward Participation in GVCs_1997 Exchange Rate Volatility Forward Participation in GVCs_1997 Exchange Rate Volatility Bacward Participation in GVCs_mean Exchange Rate Volatility Forward Participation in GVCs_mean Real Exchange Rate Bacward Participation in GVCs_t-1 Real Exchange Rate Forward Participation in GVCs_t-1 Real Exchange Rate Bacward Participation in GVCs_1997 Real Exchange Rate Forward Participation in GVCs_1997 Real Exchange Rate Bacward Participation in GVCs_mean Real Exchange Rate Forward Participation in GVCs_mean Bacward Participation in GVCs_t (0.0163) (0.0154) ** *** (0.0179) (0.0164) *** *** (0.0118) (0.0114) *** *** ( ) ( ) *** *** (0.0118) (0.0114) ** *** (0.0114) (0.0110) ( ) ** ( ) *** 0.137*** (0.0139) (0.0133) ( ) * ( ) ( ) * ( ) ( )( )( ) Forward Participation in GVCs_t ** 0.114*** (0.0179) (0.0115) Lag of Real Exports x x x x x x Real GDP of Importer x x x x x x Real GDP of Exporter x x x x x x Year Effects x x x x x x Exporter-Importer-Sector Effects x x x x x x Serial correlation First-order Second-order No.observations Note: Robust standard errors in parentheses, * 10% significance level; ** 5% significance level; ***1% significance level. 23

24 Table 7 Trade Effect of Exchange Rate Volatility Conditional on GVC Participation by Using Alternative Measure of Exchange Rate Volatility (1) (2) (3) (4) Volatility Volatility Volatility Volatility TwoYear and Current Year Two-Year Three-Year Forward Exchange Rate Volatility *** *** *** *** Exchange Rate Volatility Participation in GVCs(PGVC1_t-1 ) ( ) ( ) ( ) ( ) * *** *** ( ) ( ) ( ) ( ) Exchange Rate Volatility *** *** *** *** ( ) ( ) ( ) ( ) Exchange Rate Volatility Participation in ** *** *** *** GVCs(PGVC1_1997) ( ) ( ) ( ) ( ) Exchange Rate Volatility *** *** *** *** ( ) ( ) ( ) ( ) Exchange Rate Volatility Participation in ** *** *** GVCs(PGVC1_mean) ( ) ( ) ( ) ( ) Exchange Rate Volatility *** *** *** ( ) ( ) ( ) ( ) Exchange Rate Volatility Participation in ** GVCs(PGVC2_t-1 ) ( ) (0.0118) (0.0104) (0.0129) Exchange Rate Volatility *** *** *** * ( ) ( ) ( ) ( ) Exchange Rate Volatility Participation in * ** *** GVCs(PGVC2_1997) ( ) (0.0114) ( ) (0.0112) Exchange Rate Volatility *** *** *** * ( ) ( ) ( ) ( ) Exchange Rate Volatility Participation in * *** *** GVCs(PGVC2_mean) ( ) (0.0117) ( ) (0.0111) Real Exchange Rate x x x x Real Exchange Rate Participation in GVCs x x x x Lag of Real Exports x x x x Real GDP of Importer x x x x Real GDP of Exporter x x x x Year Effects x x x x Exporter-Importer-Sector Effects x x x x Note: Robust standard errors in parentheses, * 10% significance level; ** 5% significance level; ***1% significance level. 24