Impacts of exchange rate on Vietnam Japan trade balance: A nonlinear asymmetric cointegration approach

Nội dung text: Impacts of exchange rate on Vietnam Japan trade balance: A nonlinear asymmetric cointegration approach

1. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 IMPACTS OF EXCHANGE RATE ON VIETNAM JAPAN TRADE BALANCE: A NONLINEAR ASYMMETRIC COINTEGRATION APPROACH Tran Thi Ha 1, Nguyễn Quang Dong2, Nguyen Duy Dat3 1)National Economic University, Email: thienha1111@gmail.com 2)National Economic University, Email: dongnqneu@gmail.com 3)Thuong Mai University, Email: duydatvcu@gmail.com ABSTRACT The paper examines the impacts of exchange rate on Vietnam‘s trade balance with Japan based on the employment of industry-level data in a set of linear and nonlinear auto-regressive distributed lag models. Re- sults from the models indicate a degree of bias in regression when using aggregate data and a linear ARDL approach. Among 19 industries un- der consideration, the NARDL model presents different responses from 16 industries, which account for 40% of imports and 60% of exports between Vietnam and Japan, to exchange rate movements. The model using aggregate data shows that exchange rate positively affects Vi- etnam-Japan trade balance in case of currency depreciation, whereas currency appreciation has no impact on the trade balance between the two countries. Key word: Exchange rate, trade balance, ARDL, NARDL 1. Introduction Japan is one of Vietnam‘s most important trade partners. The market accounts for about 10% of total export-import turnover and ranks fourth among the top trading partners of Vietnam (Data from the General De- 39
2. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 partment of Vietnam Customs). It is noticeable that Vietnam has main- tained a relatively balanced trade relationship with Japan in recent years. Japan is considered a potential market that Vietnam has not been able to fully exploit its advantages – given all the economic and diplo- matic ties established by both countries, especially after the signing of the Vietnam-Japan economic partnership agreement on 25th December 2008, which later came into force on 1st October 2009. In fact, this is regarded Vietnam‘s first bilateral free trade agreement (FTA), with both countries giving much more preferential treatments for mutual trade and economic partnership as compared with the ASEAN-Japan FTA. Figure 1: Vietnam-Japan bilateral trade turnover 1995-2017 (million USD) Source: General Statistics Office and author‘s calculation Figure 1 shows that total trade turnover between Vietnam and Japan continuously increased in the past 20 years, especially after the eco- nomic partnership agreement between two countries was signed in 2008 with various preferential treatments granted to each other. Both the 40
3. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 lines have positive slopes, highlighting the fact that, from 2009 to 2017, Vietnam‘s exports to Japan increased by 160% while imports also ex- perienced a lower growth rate of 140%. Regarding the trade balance, between 1995 and 2017, Vietnam achieved trade surplus for most of the period, except for 2009-2010 and the most recent three years 2015- 2017. Therefore, it is evident that after the signing of this economic partnership agreement, bilateral trade value has while exports and imports were spotted to rise at different rates - though the differ- ence in speed was rather small. During the last three years, imports have outpaced exports, causing a trade deficit in Vietnam‘s bilateral trade relations with Japan. In particular, imports from Vietnam just accounted for 2.42 % of market share in the Japanese market in 2016 (Figure 2) while only 6.7% of Vietnam‘s exports were deliv- ered to Japan. The chart also illustrates that though the proportion of Vietnam‘s goods exported to Japan has increased in recent years, the role of the Japanese market as a Vietnam‘s major trading part- ner steadily declined. Indeed, from a moderate market share of 11.45% in 2011, Vietnam‘s goods exported to Japan dropped to a low of 6.7% in 2016. In general, these figures demonstrate that Vi- etnam has failed to take full advantage of the potential benefits brought by the FTA. A lot of experts, though with different perspectives, have suggested that Vietnam‘s growing trade deficit in recent years (2015-2017) stemmed from the appreciation of the Vietnamese Dong (VND) against the Yen (JPY). However, few quantitative studies have been conducted to support this argument so far. For that reason, this paper aims to an- 41
4. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 swer the question of how exchange rate affects Vietnam-Japan trade balance and whether or not exchange rate is an effective tool to improve Vietnam‘s trade balance with Japan. Figure 2: Share of Vietnamese goods in Japanese market (2007-2016) Source: ITC (International Trade Centre) and author‘s calculation The rest of the paper is organized as follows. Section 2 provides an overview of previous research and addresses potential research gaps. Section 3 focuses on model specification and empirical methodology, as well as steps to conduct estimation procedures for the ARDL and NARDL models. Section 4 presents and discusses the empirical results of the models. Section 5 sketches out major conclusions and policy rec- ommendations of the paper. 2. Literature review The relationship between exchange rate and trade balance has always been a highly-debated topic among academics over the past decades. However, since the middle of the twentieth century, with developments in macroeconomic and econometric analysis, several empirical studies have been carried out. Despite the emergence of more advanced re- search methods and the availability of more comprehensive databases, the relationship between exchange rate and trade balance still causes 42
5. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 much controversy and no consensus has been reached on this matter so far. In general, most studies could be divided into three groups that rep- resent three different approaches commonly adopted by scholars: i) Elasticity approach, ii), Keynesian absorption approach, and iii) Mone- tary approach. Both Keynesian absorption and monetary approach were developed with a special focus on macroeconomic linkages and identi- ties while elasticity approach is mainly adopted to study microeconom- ic relationships. Thus, the relationship between exchange rate, trade balance and other macro variables will be better clarified through the first two approaches. Still, the literature review shows that very few empirical studies have employed these two methods in their research. Most studies on the relationship between exchange rate and trade re- ly on the test of the Marshall-Lerner condition (MLC) through identify- ing the price elasticities via export and import demand functions, and more importantly, to examine the existence of the J-curve following currency devaluation. Typical studies that implemented the elasticity approach based on aggregate export and import demand functions are: Bahmani-Oskooee (1991); Arize (1987); Arize (1994); Khan (1974); Houthakker and Magee (1969); Magee (1973); Junz and Rhomberg (1973); Laffer (1977); and Bahmani-Oskooee (1985). Moreover, as confirmed by Bahmani-Oskooee and Zhang (2014), the Marshall-Lerner condition can only explore the indirect impacts of ex- change rate on trade balance via the export demand and supply func- tions. As such, many other studies have attempted to investigate the di- rect relationship between exchange rate, trade balance and other macro variables (Bahmani-Oskooee 1985, Felrningham 1988, Bahmani- Oskooee and Pourheydarian 1991, Anju and Uma 1999, Rose and Yellen 1989a). Most of them used the cointegration analysis to evaluate the relationship between variables in the long run and error-correction modeling to observe the short-term impacts of exchange rate on trade balance and the existence of the J-curve effect. Yet, these papers gave 43
6. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 conflicting conclusions on the role of exchange rate in export-import performance. Bahmani-Oskooee and Ardalani (2006) then pointed out that the use of aggregate export-import data might have led to biased results. In a recent study, Baek (2013) also indicated that studies looking at the relationship between exchange rate and trade normally have three different approaches for data usage: First, the approach based on ag- gregate trade data, or data on the exports and imports of the whole economy; Second, the approach based on aggregate trade data at the bi- lateral level; Third, the approach based on the examination of trade data at industry level. Specifically, the first approach focuses on the use of aggregate ex- port and import data between a country and the rest of the world while assessing the impacts of exchange rate depreciation or appreciation (e.g. Wilson and Takacs (1979) Bahmani-Oskooee (1986) Felmingham and Divisekera (1986) ). These studies drew different conclusions on the impacts of currency devaluation on trade balance. Meanwhile, the se- cond approach uses bilateral trade data between a country and its major trading partners to study the effects of exchange rate on trade (e.g. Rose and Yellen (1989b) Bahmani-Oskooee and Goswami (2004) Bahmani- Oskooee and Kantipong (2001) Halicioglu (2007) ). However, similar to the first approach, the authors also provided mixed (different or even opposite) conclusions on the impacts of exchange rate on trade. In fact, recent debates have suggested that such discrepancies or contradictions came from different choices of data approach, as the use of aggregate trade data has led to biased results. Since the work of Bahmani- Oskooee and Ardalani (2006), there has been a growing body of litera- ture arguing that the first and second approaches may suffer from the aggregation bias problem as exchange rate may put significant impacts on some particular industries or commodities yet might not bring about any/or just less robust effects, sometimes even in a negative way, to 44
7. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 some other commodities or industries. Hence, using aggregate trade da- ta may lead to mixed results, depending on which commodity groups are given the dominant positions. Bahmani-Oskooee and Ardalani (2006) was the first research work to address the shortcomings of the first and second approaches men- tioned above. The paper examined the relationship between the trade flows of the US and the rest of the world, and between the US and its major trading partners in different commodity groups/industries. The authors found that a depreciation of the US dollar would stimulate the exports of many commodity groups/industries for the US, while there would be no significant impacts on the imports of most industries. More recently, Baek (2013) estimated the impacts of exchange rate on bilat- eral trade between South Korea and Japan. Results show that Korea‘s exports to and imports from Japan were relatively sensitive to exchange rate in the short run, but less responsive in the long run. Another similar work of Baek (2014) on bilateral trade between South Korea and the US illustrated that Korea‘s major export industries were significantly affected by exchange rate volatility, both in the long run and the short run. Meanwhile, Korea‘s imports were found insensitive to changes in exchange rate. In addition, some papers have employed the industry-level data ap- proach to inspect exchange rate‘s direct impacts on trade balance. For example, (Ardalani and Bahmani-Oskooee 2007) investigated the direct relationship between exchange rate and trade balance (or the ratio of exports to imports) of 64 SITC industries (Standard International Trade Classification) in the US market. Results revealed that the J-curve effect only exists in 6 industries and exchange rate affects trade balance in 22 industries in the long run. Besides, most recent studies mentioned the presence of biased esti- mates in previous papers as they did not examine the asymmetric reac- tions of trade balance to exchange rate changes. These studies high- 45
8. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 lighted the limitations of those previous papers when studying the im- pacts of exchange rate on trade balance based on a weak and inadequate assumption, saying that trade balance‘s responses to changes in ex- change rate are always symmetrical whether the domestic currency ap- preciates or depreciates. Thus, numerous evidence has been found to prove that this could be a misleading assumption which causes ambi- guity in conclusions on the impacts of exchange rate on trade balance (Bahmani-Oskooee and Fariditavana 2015, 2016, Arize, Malindretos, and Igwe 2017) The employment of quantitative models to study the role and impacts of exchange rate on trade in Vietnam is increasingly drawing attention from Vietnamese academics. Pham (2012) found both short-term and long-term effects of exchange rate on trade balance in Vietnam. Shortly after the VND depreciates, trade balance will get worsen. However, it will improve after four quarters and a new equilibrium will be estab- lished after twelve quarters. The author used the autoregressive distrib- uted lag (ARDL) model to study the long-term effects and spotted the improvement of trade balance when the real exchange rate depreciates. An error correction model (ECM) was also implemented. Its results un- derscored an immediate decline in trade balance in the short run after the domestic currency depreciates. Pham and Nguyen (2013) used Vietnam‘s data series for the 1990- 2007 period to evaluate possible linkages between FDI inflows to Vi- etnam, Vietnam‘s exports and exchange rate. The authors applied the cointegration method for the panel data used in Pedroni (1999). The pa- per concluded that FDI inflows into Vietnam and Vietnam‘s exports are significantly influenced by exchange rate. In addition, exports from Vi- etnam to its major trading partners are also considerably affected by FDI. Thus, it can be affirmed that exchange rate affects exports via two channels: directly through relative commodity prices and indirectly via FDI. 46
9. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 Hoang (2013) ran a simplified VAR model to estimate the responses of trade balance to VND/USD real exchange rate shocks. The study in- dicated that there does exist a J-curve for Vietnam and its effects last for eleven months. Vu (2013) assessed the impacts of exchange rate on Vietnam‘s ex- ports to its trading partners such as the US, Korea, EU, Japan using the VECM model. The study examined the effects of exchange rate on each commodity group for each trading partner, while simultaneously investigating the impacts of the Chinese factor (in cluding Yuan ex- change rate and the scale of china's import to Vietnam's trading partner) on Vietnam‘s exports to major trading partners. Results implied that ―VND depreciation exerts positive impacts on exports. However, it is noteworthy that these impacts not only depend on the characteristics of each commodity group but also on the nature of the markets (although both factors are not mutually exclusive). For example, Korean and Jap- anese markets do not explicitly respond to exchange rate movements‖. Mai (2016) observed and estimated the impacts of exchange rate and other factors on Vietnam‘s fishery exports to the US and Japanese mar- kets. The paper used secondary data from Q1/2004 to Q4/2014 for the Ordinary Least Squares (OLS) approach. Results exposed that the real exchange rates (VND/JPY, VND/USD); Vietnam‘s fishery production output; the volume of fishery exports to countries other than the import- ing ones; income of importing countries (GDP) and seasonal factors do influence Vietnam‘s total fishery export turnover in both the US and Japanese markets. In particular, VND/USD real exchange rate positive- ly affects Vietnam‘s fishery export turnover to the US market while VND/JPY real exchange rate appears to put a negative impact on Vi- etnam‘s fishery export turnover to Japanese market. Thus, the literature review acknowledges a large number of studies which attempt to examine the impacts of exchange rate on trade (name- ly exports, imports, trade balance). Most of them focused on the im- 47
10. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 pacts of exchange rate on aggregate volume/turnover of exports/imports of the whole economy based on the elasticity approach – which is nor- mally accompanied by the adoption of the BRM model, the Marshall- Lerner condition and the J-curve effect. With respect to studies about Vietnam, apart from the works of Vu (2013) and Mai (2016), there has been no research estimating exchange rate impacts on Vietnam‘s trade with a focus on each commodity group/industry to address the short- comings of the use of aggregate trade data as mentioned above. How- ever, the work of Vu (2013) only considered the impacts of exchange rate on exports without investigating the role of exchange rate in deter- mining trade balance between Vietnam and its trading partners. The re- search paper conducted by Mai (2016), meanwhile, was limited to the case of the fishery sector‘s exports to the US and Japanese markets. Another limitation of current studies on the role of exchange rate in Vietnam‘s trade balance is that these papers only based on the conven- tional assumption that the responses of trade balance to exchange rate are always symmetrical and ignored other possibilities. To date, there has been no research that uses an asymmetric non-linear model to ex- amine asymmetric impacts of exchange rate on Vietnam's trade bal- ance. Therefore, the goal of this paper is to tackle the weaknesses and shortcomings mentioned above and to provide a more comprehensive methodology for examining the impacts of exchange rate on Vietnam- Japan bilateral trade which can effectively limit the presence of bias in regression, thereby drawing policy recommendations for Vietnam to improve the effectiveness of the monetary policy, especially with re- gard to its impacts on trade in the current period. This is significantly meaningful for policy-makers as it provides the overall picture of the role of exchange rate in trade and suggests a string of policy recommendations for the adjustment of exchange rate that is in accordance with the development goals of each industry and helps to 48
11. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 effectively control macro variables, stabilize the economy, stimulate exports and improve trade balance. 3. The models and the data 3.1. The models Derived from the study of Rose and Yellen (1989a) as well as the model applied in Bahmani-Oskooee and Baek (2016), the author pro- posed a research model to consider the impacts of exchange rate on the trade balance between Vietnam and Japan with industry-level data. It is notable that, in the study of Rose and Yellen (1989a), the au- thors developed a trade model for the two countries. The study started with an import demand function, whereby home country‘s import de- mand for goods (foreign country) depends on its actual income (foreign country) and the relative prices of imported goods. In particular, real income is expected to have a positive impact while relative prices are predicted to exert a negative impact on import demand. The import de- mand function of two trading partners will be presented as follows: và (1) Of which ( ) is the import demand of the home country (for- eign country), Y ( is the real income of home country (foreign country), ( ) is the relative price of imported goods as compared with domestically produced goods of the home country (foreign coun- try), which is measured in the currency of the home country (foreign country). Similarly, the export demand function in a perfectly competi- tive market is presented as follows: và (2) Of which ( ) is the export supply of the home country (foreign country), is the relative price of exported goods of the home country (calculated by deviding the domestic currency price of exported good, , by the domestic price level, P), is the relative price of exported 49
12. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 goods of the foreign country (calculated by dividing the domestic cur- rency of exported goods, by the foreign price level, ). The relative import price is calculated as follows: (3) Of which E is the nominal exchange rate, and RER is the real ex- change rate which is defined as: Similarly, the relative import price of the foreign country is: (4) At the equilibrium of the market, the volume of exported goods and the relative price between two countries are determined by the two fol- lowing equilibrium conditions: và (5) Thus, the real trade balance of the home country is the net export turnover measured in domestic currency divided by the domestic price level (P): (6) From the equations (1) and (6), we have a simplified equation as fol- lows: Accordingly, Rose and Yelle (1989) developed this following model in their research: (7) Of which is the trade balance of the US and its trading partner j, is the real income of the US, is the real income of the partner country, and is the real exchange rate between the USD and the currency of the partner country. 50
13. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 On the basis of the work of Rose and Yellen (1989a), Bahmani- Oskooee and Baek (2016) developed a model to examine the relation- ship between exchange rate and trade balance between South Korea and the US with an industry-based approach which uses data of each specif- ic industry. The model is presented as follows: (8) Of which denotes the volume of imports from South Korea to the US for the commodity/industry i, is the volume of exports from the US to the Korean market for the commodity/industry i. ( ) is real income of the US (South Korea), is the bilateral real ex- change rate. As the ratio is unit free, the model can be easily transformed into a log-linear model. Based on the work conducted by Bahmani-Oskooee and Baek (2016), the author built a model to research the relationship between exchange rate and trade balance of Vietnam with Japan by commodity group/industry. The model is illustrated as follows: (9) Of which represents the volume of imports to Vietnam from Ja- pan for the commodity group/industry i, represents the volume of exports from Vietnam to Japan for the commodity group/industry I. ( ) is the real income of Vietnam (Japan) which are measured by the industrial production index, is the bilateral real exchange rate between Vietnam Dong (VND) and the Japanese Yen (JPY). Accordingly, b is expected to be negative while c is predicted to be positive. REX is seen to be rising when the VND depreciates, thus d should be positive. 51
14. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 The equation (9) is applied to estimate only the long-run coefficients. To estimate short-run impacts of the variables on trade balance, espe- cially the short-run impacts of exchange rate (to consider the J-curve effect), following the approach of the paper, the equation (9) will be rewritten in the form of an error correction model as follows: (10) Pesaran, Shin, and Smith (2001) recommended a standard F-test for the hypotheses : . If the F-statistic is statistically significant, the null hypothesis will be rejected or in other words, there does exist cointegration among variables. Estimation using the non-linear ARDL model A vital assumption in both equation (9) and (10) is that all inde- pendentvariables will exert symmetrical impacts on the trade balance. With respect to the exchange rate, this assumption implies that the reac- tions of trade balance to exchange rate change in case of currency de- preciation are the same for currency appreciation. Alternatively, there are some arguments claiming that commercial producers and traders may respond differently to currency depreciation and appreciation, or, exchange rate may have asymmetrical impacts on trade balance. In or- der to test this hypothesis, studies of Bahmani-Oskooee and Fariditavana (2016) or Arize, Malindretos, and Igwe (2017) have in- cluded new variables to describe the movements of the variable LNREX which originally denotes logarit of real bilateral exchange rate. Of which, the variable NEG specifies the decrease of the exchange rate. In this paper, it denotes the appreciation of the VND, and the variable 52
15. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 POS signifies the increase of the exchange rate or the depreciation of the VND. Specifically, POS and NEG are defined as follows: Shin, Yu, and Greenwood-Nimmo (2014) suggested to replace the variable lnREX in the error correction model (9) by the two new varia- bles POS and NEG. Thus, the equation (9) will be rewritten and named NARDL or the non-linear autoregressive distributed lag model. (11) The authors proved that the bounds F-test employed in the ARDL model might be used in the NARDL model in a similar way. If the es- timation coefficients and have the same signs and equal in mag- nitude, it means exchange rate have symmetrical impacts on trade bal- ance in the long run. In contrast, if the two coefficients have opposite signs and different values, a firm conclusion about the existence of an asymmetrical correlation between exchange rate and trade balance can be drawn. In order to test the presence of asymmetrical impacts, the au- thors relied on the Wald test with the null hypothesis : . Sim- ilarly, short-run asymmetrical impacts will be tested based on the hy- pothesis : 53
16. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 3.2. Data resources In this study, the monthly trade data over the period 2005M1- 2017M12 are drawn from the UN Comtrade, and based on one-digit Harmonized System, we devided Vietnam's trade into 21 industries (see Appendix table 3). Data on real income (proxied by industrial production index), price level and exchange rate are collected from the International Financial Statistics database of the International Monetary Fund. Real exchange rate is defined as follow: In which NER is nominal exchange rate between VND and JPY, P ( is the price level of Vietnam (Japan). 4. Estimation results of the models 4.1. Diagnostic tests The adoption of unit root tests is to guarantee that all independent variables will be stationary at I(0) or I(1 ) or mixed, but not the I(2) and dependent variables must stationary in I (1) – which is an indispensable condition for the estimation of the ARDL model. The testing for sta- tionarity will be performed through a string of unit root tests for varia- bles included in the equation (1). The study used the standard Augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) tests. The results showed that most of the inde- pendent variables are stationary of order zero or order one according to the ADF standard with a 1% significant level and PP standard with a 5% significant level. The dependent varables are found to be not sta- tionary of order zero, satisfying the necessary conditions of the ARDL model. Due to the large number of models in this paper, the results of unit root tests are not presented in detail. The selection of the optimal lag length for the model (10) and (11) will base on the value of the AIC standards obtained from the unlimited 54
17. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 estimation of the ARDL models. On the basis of the comparison of the- se standards, the optimal lag length for the research models will be de- termined determined accordingly, first starting with a maximum of 6 lags. The results of the models with optimal lag length will be presented below and in the appendix. Results of the bounds test for each model will be illustrated in detail in the table 4 and table 5 (see appendix). In general, the results are as follows: For the ARDL models: F-statistic revealed that there are 12/19 commodity groups/industries that witness a long-term relationship be- tween variables included in the model. 6 groups/industries do not detect the presence of cointegration. For the NARDL models: F-statistic showed that there are 16/19 commodity groups/industries that can spot a long-term relationship be- tween variables embedded in the model. 6 groups do not find the pres- ence of cointegration. On the other hand, Bahmani-Oskooee and Ardalani (2006) suggested that the results of the F-bounds test are very sensitive to the number of lags chosen for the variables. It is also important to remind that Banerjee, Dolado, and Mestre (1998) and Banerjee, Dolado, and Mestre (1998) used the standards of the error-correction term ( ) to con- sider the existence of a long-term relationship between variables in the model. Accordingly, if the error correction term is negative and statisti- cally significant, it is possible to conclude that there really exists a rela- tionship between variables in the long term. Results of the error correc- tion model are presented in the Table 4 and 5 (see appendix), showing that all the error correction terms are negative and statistically signifi- cant in every model. Thus, it is possible to conclude that there does ex- ist a long-term relationship between variables in every estimation mod- el below, which satisfies the conditions for a further analysis of these models. 55
18. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 Furthermore, the paper also performed a set of model diagnostic tests including: a test for autocorrelation (LM test), a test for heteroscedastic- ity (HK test), a general specification test for the linear regression model (RESET test), a test for parameter stability (CUSUM and CUSUMSQ test), and the Wald test to check whether an asymmetric response exists in the short term and long term. Results of these tests are presented in the Table 4 (for the ARDL models) and Table 5 (for the NARDL mod- els). The results also confirmed that there is no existence of autocorrela- tion, heteroscedasticity in the models and the models are correctly spec- ified. The CUSUM and CUSUMSQ tests showed constancy of the co- efficients in most models (stable coefficients in both tests are denoted by "S" while unstable one called by "US"). Regarding the Wald test, its results confirms the presence of long-run asymmetrical impacts in 13 commodity groups and short-run effects in 12 commodity groups 4.2. Estimation results of long-run coefficients in the ARDL models The estimation results of long-run coefficients in the ARDL models are presented in the Table 1. The signs of the long-run coefficients vary among different commodity groups. The estimation results of the long-run coefficients included in the models showed that there are only 6 commodity groups/industries (ac- counting for 28% of imports and 34% of exports of Vietnam for this market in 2017) which are affected by exchange rate, including: vege- table products (S02), mineral product (S05), leather and leather prod- ucts (S08), textile materials and textile products (S11), stone and glass products (S13), base metals and metal products (S15), works of art, col- lectors‘ pieces and antiques (S21). Most commodity groups/ industries receive negative signs, except for the group of artworks (S21). The in- come coefficient of Vietnam suggested there are 6 commodity groups/industries (making up 32% of imports and 31% of exports in the Japanese market in 2017) which are statistically significant, including: live animals and products from live animals (SC02); textile materials 56
19. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 and textile products (SC11); stone, cement and glass products (SC13); base metals and metal products (SC15); optical, photographic, cine- matographic tools and equipment (SC18); works of art, collectors‘ pieces and antiques (SC21). Of which, the commodity groups/industries that have positive signs are the SC11, SC15 and SC21. As for those concluded to be statistically significant by the income coefficient of Vi- etnam, they – including leather and leather products (SC08), wood and products of wood (SC09), stone and plaster products (SC13), means of transport (SC17), and miscellaneous manufactured articles (SC20) – only occupy 10% of imports and 18% of exports of Vietnam in the Jap- anese market. Table 1: Estimation results of long-run coefficients in the ARDL models Product code Log(REX) Log(INDVN) Log(INDJP) Japan SC01 0.7964 (0.7603) -0.2954 (0.5882) 0.4325 (2.0300) SC02 -2.6837 (0.7685) -1.1848 (0.5717) -1.0245 (2.1711) SC04 2.1356 (1.5715) -0.8336 (0.7232) 4.8088 (4.1626) SC05 2.5181* (1.4521) 0.3792 (1.0631) -2.1082 (4.0783) SC06 -0.4224 (0.5138) -0.3245 (0.4101) 2.2234 (1.8443) SC07 -0.7493 (0.6074) 0.0873 (0.4362) 3.0354 (2.5639) SC08 -2.4469 (0.5936) -0.0564 (0.4664) 0.3780 (1.5536) SC09 -1.0060 (0.6751) -0.6522 (0.5382) 5.7116 (2.7731) SC10 0.6096 (0.3734) 0.0238 (0.2879) -1.4573 (0.9963) SC11 -0.5768 (0.1792) 0.3379 (0.1359) -0.8453 (0.5476) SC12 -0.8668 (0.5671) -0.3175 (0.4711) -1.3831 (1.4868) 1.8254 SC13 -1.4528 (0.1352) -0.7282 (0.1019) * (0.3847) SC14 0.0318 (0.2696) 0.5409 (0.2164) 0.5104 (0.6855) SC15 -0.6854 (0.2263) 0.7305 (0.1742) 0.6857 (0.6466) SC16 0.1041 (0.1933) 0.1602 (0.1673) -0.3780 (0.5120) 57
20. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 SC17 -0.1579 (0.6940) -0.0678 (0.5293) 4.3949 (2.1895) SC18 -0.2416 (0.3245) -0.6376 (0.3020) -1.3835 (0.8729) SC20 -0.1908 (0.1849) -0.1838 (0.1413) -1.1370 (0.4960) SC21 2.0190 (0.3432) 0.7612 (0.2582) 1.1869 (1.1068) SCtotal 0.0808 (0.3714) 0.0909 (0.2678) 2.2841 (1.5085) Note: 1. *, , means the regression coefficient is statistically significant at the 1%, 5% and 10% level respectively2. The value in brackets is the standard error of the estimate of the coefficient The coefficient of the variable denoting exchange rate in the aggre- gate model (Sctotal) of Japan is not statistically significant, implying that exchange rate does not affect the total trade balance between Vi- etnam and Japan. In general, from the results of the ARDL model, the responses of trade balance between Vietnam and Japan to changes in exchange rates only appear in 6 commodity groups/industries which account for 24% of imports and 38% of exports of Vietnam in this market. However, looking at the model that measure the impacts of exchange rate on Vi- etnam‘s total trade balance (denoted by the dependent variable Sctotal) with Japan, it is possible to conclude that exchange rate is not sensitive to trade balance in the case of Vietnam. The results justified Baek (2013)‘s conclusion, saying that the use of aggregate data will cause bias in the estimates of the coefficients. Apparently, it is not reasonable to conclude how differently the exchange rate affects the trade balance of different partners and how it affects different commodity groups/industries just by looking at the results of the aggregate model. 4.3. Estimation results of short-run coefficients in the ARDL models Table 6 (see Appendix) presents the short-run coefficient estimate of the exchange rate variable in the models of the three partners. The ex- change rate is considered to have a short-term impact on a commodity group/industry‘s trade balance if at least one short-run coefficient is sta- tistically significant. 58
21. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 Results showed that there are 11 commodity groups/industries that have at least one statistically significant short-run coefficient. These groups make up 42% of imports and 48% of exports of Vietnam in this market, including SC02 (vegetable products), prepared foodstuffs and beverages (SC04), mineral product (SC05), plastics and rubber (SC07), leather and leather products (S08), chemical products (SC06), wood and products of wood (SC09), textile materials and textile products (SC11), stone and plaster products (SC13), base metals and metal prod- ucts (SC15), works of art (SC21). Most of the short-run coefficients have negative signs, excepting those of vegetable products (SC02) and works of arts (SC21). The J-curve effect The impacts of exchange rate on the trade balance of an industry/a commodity group is considered to yield a J-curve effect when the cur- rency depreciates. In particular, the trade balance will deteriorate in the short term and improve in the long term. Thus, according to the previ- ous research ((Rose and Yellen 1989a), a commodity group/industry is likely to experience the J-curve effect if the short-run coefficient of the exchange rate variable has a negative sign while the long-run coeffi- cient has a positive sign while the long-run coefficient has a positive sign. However, the estimation results for the short-run and long-run co- efficients of the exchange rate variable showed that the J-curve theory is supported by the only case of mineral products (SC05). 4.4. Estimation results of long-run coefficients in the NARDL models Table 2 presents the long-run coefficient estimates of the NARDL model. As shown in Section 1, the NARDL model aims to test trade balance‘s response to exchange rate movements in two cases: when the currency depreciates and when the currency appreciates. The theory of the asymmetric responses of trade balance to exchange rate changes suggests that producers will respond quickly to fluctuations in markets to meet export demand. However, when the currency appreciates, ex- porters will react more slowly to changes in exchange rate due to mar- ket share constraints Therefore, it is expected that the impacts of ex- change rate on trade balance in case of currency devaluation will be larger compared with the case of currency appreciation. In this study, NEG or a negative value of ―∆LER‖ denotes the appre- ciation of the VND while POS specifies the depreciation of the VND. 59
22. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 Asymmetric reactions will not occur if the coefficients of NEG and POS have the same sign and equal in magnitude. Table 2: Estimation results of long-run coefficients in the NARDL models Product code POS NEG INDVN INDCN Japan SC01 -1.3819 (0.8833) 0.7183* (0.4258) 1.2183 (0.6129) -0.3472 (1.1886) SC02 -1.7599 (1.8807) -2.4403 (0.8319) -1.3130 (1.1013) 1.1542 (2.7931) SC04 4.8406 (2.0263) 1.8820* (0.9928) -2.6498 (0.9370) 0.9816 (2.3449) SC05 5.5089* (3.2712) 2.4853 (1.5754) -1.6377 (1.9831) 4.1585 (4.6704) SC06 0.5406 (0.6328) -0.3279 (0.2917) -0.8015 (0.3792) 2.5403 (1.2008) SC07 0.6520 (0.4323) -0.8183 (0.1854) -0.8350 (0.2605) 1.3210* (0.6867) SC08 -3.1442 (1.3367) -2.4172 (0.6326) 0.2535 (0.8206) -0.3229 (1.6184) SC09 0.8367 (1.2116) -1.0388* (0.5574) -1.8273 (0.7585) 5.2695 (2.2727) SC10 1.6186 (0.5997) 0.6145 (0.2783) -0.7077* (0.3584) -0.7886 (0.7610) SC11 0.1279 (0.1023) -0.5706 (0.0509) -0.1585 (0.0717) -0.1381 (0.1203) SC12 1.7006 (0.3821) -0.3682 (0.1770) -1.1905 (0.2275) -0.3968 (0.4855) SC13 -1.0499 (0.2980) -1.4373 (0.1330) -1.0095 (0.1843) 2.0154 (0.3950) SC14 -1.0599* (0.5960) -0.0387 (0.2823) 1.2108 (0.4038) -0.2790 (0.8412) SC15 0.4388 (0.3835) -0.6822 (0.1766) -0.0966 (0.2196) 1.5682 (0.5650) SC16 0.1964 (0.4283) 0.1183 (0.1966) 0.1257 (0.3251) -0.4802 (0.5596) SC17 0.1871 (1.9038) 0.0722 (0.8300) 0.3208 (1.1378) 7.6848 (3.6990) SC18 0.9102 (0.3816) -0.0567 (0.1824) -1.0795 (0.2307) -0.5476 (0.4830) SC20 -0.8942 (0.2773) -0.2375* (0.1276) 0.2469 (0.1733) -1.1858 (0.3518) SC21 2.9854 (0.5978) 2.0845 (0.2668) 0.1272 (0.3464) 1.5569* (0.8813) SCtotal 1.2032* (0.7186) 0.1449 (0.3017) -0.5778 (0.4037) 1.9493* (1.1385) Source: Author‘s calculations. Note: 1. * Indicates significance at the 10% level , at the 5% level, and at the 1% level 2. Numbers inside parentheses are the standard deviation of the coefficients Results of the NARDL models pointed out that there are 16 com- modity groups/ industries (accounting for 46% of imports and 63% of 60
23. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 exports between Vietnam and Japan) which are sensitive to exchange rate changes including 7 commodity groups/industries that are specified by the ARDL models: vegetable products (S02), mineral products (SC05), leather and leather products (S08), textile materials and textile products (S11), stone and glass products (S13), base metals and metal products (S15), works of art, collectors‘ pieces and antiques (S21). 9 other commodity groups/industries pointed out by the NARDL models include: live animals and products from live animals (SC01), prepared foodstuffs and beverages (SC04), plastic and plastic products (SC07), paper and paper products (SC10), footwear (SC12), precious stones and precious metals (SC14), optical, photographic, cinematographic tools and equipment (SC18), miscellaneous manufactured articles (SC20), works of art, collectors‘ pieces and antiques (SC21). There is a notable difference between the ARDL model and the NARDL model. In the NARDL model, trade balance is found to be more sensitive to exchange rate movements than in the ARDL model. This finding is consistent with many previous studies, pointing at the problem of bias that en- counters previous works when assuming that the impacts of exchange rate on trade balance is always symmetrical (Bahmani-Oskooee and Baek, 2016; Arize et al., 2017). Clearly, without the adoption of the non-linear regression model (NARDL), these commodity groups/industries might be mistaken that their real exchange rate coeffi- cients were not statistically significant. Or in other words, their trade balance did not respond to the change of exchange rate. The coefficient results of SC01 showed that exchange rate only af- fects the trade balance of the group. When the VND appreciates by 1 per cent against the JPY, the trade balance deteriorates by 0.72%. SC02, SC07, SC09, SC11 and SC15 also send out asymmetric respons- es to the appreciation of the VND but the trade balance of these indus- tries improve by 0.5-2.4%. In case of VND devaluation, the trade bal- ance of SC05 and SC18 respond positively to the change of exchange 61
24. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 rate. If, the real exchange rate RER increases by 1%, the trade balance of these industries will increase by 0.9-5.5%. The SC04, SC10 and SC21 commodity groups/industries are found to have symmetrical reac- tions, i.e. if exchange rate increases (or VND depreciation), trade bal- ance will improve by 1.6-4.8%; if exchange rate decreases (or VND appreciation) trade balance will get worsened by 0.6-2%. In contrast, the SC08, and SC13 groups also show symmetrical responses but their trade balance become worsened when the VND depreciates and im- prove when the VND appreciates. Athought SC20 also have same posi- tive sign for NEG and POS, but the Wald test (see table 5) suggests that the deprecitation and appreciation of the VND will exert long-run asymmetrical effects on trade balance. The SC12 group reacts to both the devaluation and the appreciation of the VND and the trade balance improves in both cases. However, the absolute magnitude of the NEG and POS coefficients implies that trade balance might have a stronger reaction to exchange rate in case of currency appreciation. Concerning the aggregate model with the variable SCtotal denoting product code, the results indicated that exchange rate does affect the trade balance between Vietnam and Japan in case of VND depreciation. If the VND depreciates by 1%, Vietnam‘s trade balance with Japan will improve by 1.2%. In case of VND appreciation, the trade balance is not affected. In general, the results of the NARDL model have again emphasized the limitations of the ARDL model‘s approach with its one-size-fits-all assumption saying that the effects of exchange rate on trade balance are always symmetrical. Results showed that the impacts of exchange rate are found to be asymmetric in many commodity groups/industries. The conclusions of the NARDL model confirmed that a lot of commodity groups/industries have witnessed the effects of exchange rate on trade balance with statistically significant POS and NEG coefficients. Mean- while, if we looked at the results yielded from the ARDL model, these 62
25. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 commodity groups/industries are found to not respond to exchange rate changes. Among the 16 commodity groups/industries that have sensi- tive trade balance to changes in exchange rate, 9 groups/industries pro- duce asymmetric responses. On a side note, the devaluation of the VND does not necessarily mean that the trade balance will improve. Only 6 out of 10 groups/industries have positive POS coefficients or the trade balance will improve when the VND depreciates. In case of VND ap- preciation, there are 13 affected commodity groups/industries (as their coefficients of NEG are statistically significant). Only 4 groups/industries suffer from deteriorating trade balance. In contrast, the results of the model run on aggregate trade data showed that ex- change rate only affects trade balance in case of VND depreciation and benefits Vietnam as it will improve the country‘s trade balance. 4.5. Estimation results of short-run coefficients in the NARDL models Similarly, the short-run coefficient estimates of the NARDL models are presented in Table 7 (See Appendix). The results of the Wald test statistic (See table 5 as Wald_S) indicate that we can reject the null hy- pothesis of symmetry in the short-run for 11 commodity groups. Of which, commodity groups/industries that are sensitive to exchange rate when the VND depreciates include: SC02, SC08, SC10, SC12, SC18 and SC20. Meanwhile, SC05, SC06, SC10, SC11, SC12, SC15, SC18, and SC20 are sensitive to exchange rate when the VND appreciates. Overall, considering the estimation results of the short-term coefficients included in the NARDL models, it is apparent that using the NARDL model can detect more commodity groups/industries that respond to ex- change rate than using the ARDL model. The J-curve effect In the NARDL model, an industry is considered to have the J-curve effect when there is at least a negative short-term coefficient, provided that the short-term coefficients at the previous lag are not positive and statistically significant – or all coefficients are not statistically signifi- 63
26. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 cant – plus the long-term coefficient of POS is positive and statistically significant. According to the results of the models, the J-curve effect is found in SC04, SC05 and SC21. However, the J-curve effect is not spotted in the Sctotal aggregate model. In summary, the results of the models show that not much evidence was found in the case of trade relations between Vietnam and Japan to support the theory of the J-curve effect. 5. Conclusion and policy recommendations The paper exclusively employs the linear ARDL and non-linear ARDL (NARDL) models to examine the trade relations between Vi- etnam and Japan in each industry (21 industries based on the classifica- tion of goods specified the Harmonized System – HS). In particular, the paper seeks to overcome the shortcomings of previous works in as- sessing the relationship between trade balance and exchange rate in Vi- etnam by: (1) using industry data to address limitations raised by the utilisation of aggregate data, and (2) employing a non-linear model to avoid bias when assuming that trade balance produces symmetrical re- sponses after the currency depreciates or appreciates. Given the estimation results of the short-term and long-term coeffi- cients included in the model, several conclusions can be drawn as fol- lows: First, the estimation results of the ARDL and NARDL models for all industries revealed a degree of bias to the regression estimates when using aggregate data for the asymmetric approach. Therefore, the au- thor‘s approach of using the NARLD model with industry-level data would provide a more comprehensive and complete picture of the ef- fects of exchange rate changes on Vietnam‘s trade balance. Second, the theory on the J-curve effect was not supported by the es- timation results of the model examining the relationship between Vi- etnam-Japan trade balance and the exchange rate. Results from models using industry-specific data implied that only few industries have expe- 64
27. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 rienced the J-curve effect while the models employing aggregate data did not find any signs of the J-curve effect. Third, Japan‘s income (proxied by the industrial production index) had a positive impact on Vietnam-Japan trade balance, whereas Vi- etnam‘s income did not. Fourth, the paper shows that trade balance of all industries are rela- tively sensitive to exchange rate (16/19 industries). Furthermore, the industries‘ trade balance respond differently to exchange rate‘s changes. However, the industries that show significant responses do not contrib- ute a large proportion in bilateral trade turnover, in fact only affecting roughly 60% of imports and 40% of exports between Vietnam and Ja- pan. Results from the NARDL models with aggregate data indicated that, when VND depreciates, Vietnam‘s trade balance with Japan will improve. In the case where VND appreciates, the trade balance does not respond to changes in exchange rate. Therefore, the exchange rate seems to be an effective tool in improving Vietnam‘s trade balance in the Japanese market. Thus, results implied that exchange rate really affects the trade bal- ance between Vietnam and Japan. Nevertheless, the magnitude of such impacts depend on each industry. Given the conclusions above, it is plausible to consider exchange rate an effective tool to stimulate ex- ports and improve Vietnam‘s trade balance. However, the use of ex- change rate should be well-thought-out based on different priority tar- gets and the coordination between different policies. For example, in view of targets relating to inflation, foreign debts or the stabilization of the currency market, a depreciation of VND is expected to exert nega- tive effects on these targets. 65
28. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 REFERCENCES Anju, Gupta-Kapoor, and Ramakrishnan Uma. 1999. "Is there a J- curve? A new estimation for Japan." International Economic Journal 13 (4):71-79. Ardalani, Zohre, and Mohsen Bahmani-Oskooee. 2007. "Is there a J- Curve at the Industry Level?" Economics Bulletin 6 (26):1-12. Arize, Augustine. 1987. "The supply and demand for imports and exports in a simultaneous model." Applied Economics 19 (9):1233-1247. Arize, Augustine C. 1994. "Cointegration test of a long-run relation between the real effective exchange rate and the trade balance." International Economic Journal 8 (3):1-9. Arize, Augustine C, John Malindretos, and Emmanuel U Igwe. 2017. "Do exchange rate changes improve the trade balance: An asymmetric nonlinear cointegration approach." International Review of Economics & Finance 49:313-326. Baek, Jungho. 2013. "Does the exchange rate matter to bilateral trade between Korea and Japan? Evidence from commodity trade data." Economic Modelling 30:856-862. doi: Baek, Jungho. 2014. "Exchange rate effects on Korea–U.S. bilateral trade: A new look." Research in Economics 68 (3):214-221. doi: Bahmani-Oskooee, Mohsen. 1985. "Devaluation and the J-curve: some evidence from LDCs." The review of Economics and Statistics:500-504. Bahmani-Oskooee, Mohsen. 1986. "Determinants of international trade flows: the case of developing countries." Journal of Development Economics 20 (1):107-123. 66
29. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 Bahmani-Oskooee, Mohsen. 1991. "Is there a long-run relation between the trade balance and the real effective exchange rate of LDCs?" Economics letters 36 (4):403-407. Bahmani-Oskooee, Mohsen, and Zohre Ardalani. 2006. "Exchange Rate Sensitivity of U.S. Trade Flows: Evidence from Industry Data." Southern Economic Journal 72 (3):542. Bahmani-Oskooee, Mohsen, and Jungho Baek. 2016. "Do exchange rate changes have symmetric or asymmetric effects on the trade balance? Evidence from US–Korea commodity trade." Journal of Asian Economics 45:15-30. Bahmani-Oskooee, Mohsen, and Hadise Fariditavana. 2015. "Nonlinear ARDL approach, asymmetric effects and the J- curve." Journal of Economic Studies 42 (3):519-530. Bahmani-Oskooee, Mohsen, and Hadise Fariditavana. 2016. "Nonlinear ARDL approach and the J-curve phenomenon." Open Economies Review 27 (1):51-70. Bahmani-Oskooee, Mohsen, and Gour Gobinda Goswami. 2004. "Exchange Rate Sensitivity of Japan's Bilateral Trade Flows." Japan and the World Economy 16 (2004) 1-15. Bahmani-Oskooee, Mohsen, and Tatchawan Kantipong. 2001. "Bilateral J-curve between Thailand and her trading partners." Journal of Economic Development 26 (2):107-118. Bahmani-Oskooee, Mohsen, and Mohammad Pourheydarian. 1991. "The Australian J-curve: a reexamination." International Economic Journal 5 (3):49-58. Bahmani-Oskooee, Mohsen, and Ruixin Zhang. 2014. "Is there J- Curve effect in the commodity trade between Korea and rest of the world?" Economic Change and Restructuring 47 (3):227- 250. Banerjee, Anindya, Juan Dolado, and Ricardo Mestre. 1998. "Error‐ correction mechanism tests for cointegration in a 67
30. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 single‐ equation framework." Journal of time series analysis 19 (3):267-283. Felmingham, Bruce S, and Sarath Divisekera. 1986. "The response of Australia's trade balance under different exchange rate regimes." Australian Economic Papers 25 (46):33-46. Felrningham, BS. 1988. "Where is the Australian J‐ curve?" Bulletin of Economic Research 40 (1):43-56. Halicioglu, Ferda. 2007. "The J-curve dynamics of Turkish bilateral trade: a cointegration approach." Journal of Economic Studies 34 (2):103-119. Hoang, Van Khieu. 2013. "The effects of the real exchange rate on the trade balance: Is there a J-curve for Vietnam? A VAR approach.". Houthakker, Hendrik S, and Stephen P Magee. 1969. "Income and price elasticities in world trade." The Review of Economics and Statistics:111-125. Junz, Helen B, and Rudolf R Rhomberg. 1973. "Price competitiveness in export trade among industrial countries." The American Economic Review 63 (2):412-418. Khan, Mohsin S. 1974. "Import and export demand in developing countries." Staff Papers 21 (3):678-693. Laffer, Arthur B. 1977. "Exchange rates, the terms of trade, and the trade balance." Effects of exchange rate adjustments. Magee, Stephen P. 1973. "Currency contracts, pass-through, and devaluation." Brookings Papers on Economic Activity 1973 (1):303-325. Mai, Thi Cam Tu. 2016. "Impact of exchange rate on Vietnam seafood exports - The cases of Japan and US markets." Journal of Development and Integration (Vietnam) 26 (36):42-52. 68
31. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 Pedroni, Peter. 1999. "Critical values for cointegration tests in heterogeneous panels with multiple regressors." Oxford Bulletin of Economics and statistics 61 (S1):653-670. Pesaran, M Hashem, Yongcheol Shin, and Richard J Smith. 2001. "Bounds testing approaches to the analysis of level relationships." Journal of applied econometrics 16 (3):289- 326. Pham, Thi Hong Hanh, and Duc Thinh Nguyen. 2013. "Foreign direct investment, exports and real exchange rate linkages in Vietnam: evidence from a co-integration approach." Journal of Southeast Asian Economies (JSEAE) 30 (3):250-262. Pham, Thi Tuyet Trinh. 2012. "The Impact of Exchange Rate Fluctuation on Trade Balance in Short and Long Run." Depocen Working Paper Series 2012 (23). Rose, A, and J. L. Yellen. 1989a. "Is There a J-curve?" Journal of Monetary economics 24:53-68. Rose, Andrew K, and Janet L Yellen. 1989b. "Is there a J-curve?" Journal of Monetary economics 24 (1):53-68. Shin, Yongcheol, Byungchul Yu, and Matthew Greenwood-Nimmo. 2014. "Modelling asymmetric cointegration and dynamic multipliers in a nonlinear ARDL framework." In Festschrift in Honor of Peter Schmidt, 281-314. Springer. Vu, Quoc Huy. 2013. Tỷ giá hối đoái giai đoạn 2000-2011: Mức độ sai lệch và tác động đối với xuất khẩu: Nhà xuất bản tri thức trẻ. Wilson, John F, and Wendy E Takacs. 1979. "Differential responses to price and exchange rate influences in the foreign trade of selected industrial countries." The review of Economics and Statistics:267-279. 69
32. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 APPENDIX Table 3: HS Classification by section (21 sections) Product code Name of product SC01 Live animals; products from animals SC02 Vegetable products Animal or vegetable fats and oils and their cleavage products; SC03 prepared edible fats; animal or vegetable waxes Prepared foodstuffs; beverages, spirits and vinegar; tobacco and SC04 manufactured tobacco substitutes SC05 Mineral products SC06 Products of the chemical or allied industries SC07 Plastics and articles thereof; rubber and articles thereof Raw hides and skins, leather, furskins and articles thereof; sad- dlery and harness; travel goods, handbags and similar containers; SC08 articles of animal gut (other than silk-worm gut) Wood and articles of wood; wood charcoal; cork and articles of cork; manufactures of straw, of esparto or of other plaiting mate- SC09 rials; basketware and wickerwork Pulp of wood or of other fibrous cellulosic material; recovered (waste and scrap) paper or paperboard; paper and paperboard and SC10 articles thereof SC11 Textiles and textile articles Footwear, headgear, umbrellas, sun umbrellas, walking-sticks, seat-sticks, whips, riding-crops and parts thereof; prepared feath- ers and articles made therewith; artificial flowers; articles of hu- SC12 man hair 70
33. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 Articles of stone, plaster, cement, asbestos, mica or similar mate- SC13 rials; ceramic products; glass and glassware Natural or cultured pearls, precious or semi-precious stones, pre- cious metals, metals clad with precious metal and articles there- SC14 of; imitation jewellery; coin SC15 Base metals and articles of base metal Machinery and mechanical appliances; electrical equipment; parts thereof; sound recorders and reproducers, television image and sound recorders and reproducers, and parts and accessories SC16 of such articles SC17 Vehicles, aircraft, vessels and associated Optical, photographic, cinematographic, measuring, checking, precision, medical or surgical instruments and apparatus; clocks SC18 and watches; musical instruments; parts and accessories thereof SC19 Arms and ammunition; parts and accessories thereof SC20 Miscellaneous manufactured articles SC21 Works of art, collectors‘ pieces and antiques 71
34. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 Table 4: Results of the bound tests and diagnostic tests for the ARDL models Product F ECM(t-1) LM RESET HSK test CUSUM CUSUMSQ R2 Japan sc01 5.502 -0.3265 6.836* 3.828 15.23 S S 0.3186 sc02 5.945 -0.3203 0.397 0.455 4.997 S S 0.3286 sc04 4.306* -0.1577* 0.833 0.458 1.713 S S 0.4951 sc05 4.007* -0.4125 3.22 1.192 0.884 S S 0.4098 sc06 2.764 -0.2110 2.366 2.944 0.105 S US 0.4907 sc07 1.573 -0.1310 2.67 0.666 0.0898 S US 0.5341 sc08 5.876 -0.2975 4.613 1.653 3.499* S S 0.2137 sc09 4.122* -0.3384 2.663 0.431 5.052* S S 0.6923 sc10 3.93* -0.2714 3.637 0.295 0.231 S US 0.3532 sc11 2.616 -0.4016 3.649 2.967 2.969 S US 0.6603 sc12 3.047 -0.3777 4.686 1.160 0.543 US S 0.5351 sc13 11.336 -1.2724 3.644 1.115 1.043 S S 0.5253 sc14 24.126 -0.8611 22.13 1.829 86.42 S US 0.4727 sc15 10.699 -0.4938 0.577 3.943 0.0359 S S 0.3016 sc16 3.521 -0.4763 3.22 2.400* 0.152 S US 0.5666 sc17 3.610 -0.1950 1.249 0.370 0.942 S S 0.2991 sc18 5.952 -0.2912 4.065 1.413 0.123 S S 0.5246 sc20 3.737* -0.4574 2.247 0.724 0.997 S S 0.5706 sc21 8.786 -0.5652 3.634 3.311 0.189 S S 0.4331 Total 2.895 -0.2078 4.553 1.512 3.699* S S 0.5524 Note: 1. The upper bound critical value of the F-test for cointegration is 3,52 for the 10% level of significance and 5.06 for the 1% level. See Pesaran et al. (2001, Table CI, CaseV, p. 301.) 2. * Indicates significance at the 10% level , at the 5% level, and at the 1% level 72
35. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 Table 5: Results of the bound tests and diagnostic tests for the NARDL models Product F ECM(t-1) LM RESET HSK test Wald_l Wald_s CUSUM CUSUMSQ R2 sc01 5.746 -0.4577 2.205 0.52 2.796 8.41 0.21 S US 0.396 sc02 5.194 -0.3035 2.715 1.442 2.137 0.20 3.92 S US 0.3698 sc04 4.319 -0.2204 2.193 0.195 0.553 4.95 0.31 S S 0.5835 sc05 4.417 -0.4686 5.758 1.404 2.605 1.38 4.78 S S 0.5339 sc06 5.445 -0.3469 3.133 2.136 1.132 2.91* 6.49 S US 0.6051 sc07 5.354 -0.3876 6.439 2.372 0.161 18.22 2.28 0.5977 sc08 6.499 -0.3359 6.311 1.547 0.126 0.48 9.56 S US 0.2653 sc09 2.84 -0.4311 6.392 1.343 1.528 3.7* 0.36 S US 0.6454 sc10 4.403 -0.3642 2.973 0.671 2.502 4.43 3.35* S S 0.3721 sc11 42.647 -1.2895 3.233 0.0811 0.0892 69.47 53.94 S NS 0.6828 sc12 27.425 -1.3298 1.213 2.253 0.0293 46.49 32.02 US US 0.5617 sc13 9.691 -1.1350 5.984 5.059 4.536 2.57 2.64 S US 0.5411 sc14 13.216 -0.9884 3.971 3.593 33.63 4.42 0.19 S US 0.6472 sc15 10.602 -0.6418 1.667 1.169 0.624 14.10 3.45* S S 0.4325 sc16 2.812 -0.4725 3.14 0.846 0.145 0.04 0.05 S US 0.5668 sc17 2.534 -0.1704 3.729 3.82* 0.0825 0.006 8.47 S S 0.4432 sc18 6.345 -0.5219 4.761 2.561 3.439 10.92 4.82 S S 0.5803 sc20 6.434 -0.6629 7.104 1.988 0.991 8.58 7.28 S S 0.5893 sc21 13.484 -0.7344 2.813 1.889 0.316 3.58* 1.32 s S 0.4586 Total 2.684 -0.2618 3.558 1.729 0.27 3.59* 4.19 S S 0.569 Note: 1. The upper bound critical value of the F-test for cointegration is 3.52 for the 10% level of significance and 5.06 for the 1% level. See Pesaran et al. (2001, Table CI, CaseV, p. 301.) 2. * Indicates significance at the 10% level , at the 5% level, and at the 1% level 73
36. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 Table 6: Estimation of short-run coefficient in the ARDL models Product code Lag 0 Lag 1 Lag 2 Lag 3 Lag 4 Lag 5 Constant Observations R-squared SC01 0.20 (0.21) -0.78 (2.61) 114 0.34 SC02 1.54 (1.35) -2.58* (1.39) 1.25 (1.40) -1.07 (1.39) 3.25 (1.36) -2.17 (1.35) 8.39 (4.06) 114 0.33 SC04 0.11 (0.87) -1.14 (0.89) 0.30 (0.91) -0.99 (0.89) -2.36 (0.89) -4.26* (2.34) 114 0.50 SC05 4.05 (3.90) -2.25 (4.01) -2.75 (4.08) -7.48* (3.97) -10.77 (3.99) -5.54 (4.04) -1.12 (10.41) 114 0.53 SC06 0.11 (0.64) -1.34 (0.64) -1.44 (1.59) 114 0.49 SC07 0.73 (0.45) -0.12 (0.43) 0.38 (0.43) -0.73* (0.44) -0.98 (0.43) -1.45 (1.16) 114 0.53 SC08 1.43 (1.08) 1.68 (1.12) 1.89 (1.15) 0.16 (1.14) -0.08 (1.10) 2.87 (1.08) 4.86* (2.71) 114 0.30 SC09 -0.77 (1.29) -2.88 (1.32) -0.01 (1.35) -1.41 (1.33) 3.21 (1.32) -3.93 (1.30) -4.47 (3.59) 114 0.69 SC10 0.17 (0.11) 0.78 (1.40) 114 0.35 SC11 -0.72* (0.41) -0.96 (0.42) 2.59 (1.10) 114 0.66 SC12 -0.33 (0.22) 6.73 (3.18) 114 0.54 SC13 -1.85 (0.33) 3.76 (2.61) 114 0.53 SC14 0.22 (1.55) -4.71 (3.83) 116 0.60 SC15 -0.34 (0.13) -2.41 (1.64) 114 0.30 SC16 0.05 (0.10) 0.06 (1.30) 114 0.57 SC17 -0.03 (0.14) -3.78* (2.13) 114 0.30 SC18 -0.07 (0.09) 3.00 (1.24) 114 0.52 SC20 -0.09 (0.09) 4.11 (1.49) 114 0.57 SC21 1.14 (0.26) -11.62 (3.32) 114 0.43 SCtotal 0.02 (0.07) -2.30 (1.10) 114 0.55 Note: 1. * Indicates significance at the 10% level , at the 5% level, and at the 1% level 2. Numbers inside parentheses are the standard deviation of the coefficients 74
37. Hội thảo Khoa học quốc gia “Thương mại quốc tế - Chính sách và thực tiễn tại Việt Nam”, ISBN: 978 - 604 - 67 - 1403 - 3 Table 7: Short-run coefficients estimates of POS in the NARDL models Lag 0 Lag 1 Lag 2 Lag 3 Lag 4 Lag 5 SC01 -1.4670 (2.5382) 1.3702 (2.60) 0.6561 (2.82) SC02 2.6979 (3.3165) -0.0996 (3.15) 7.7051 (3.40) SC04 1.9186 (1.9952) 2.9605 (1.94) -3.8541* (2.08) -1.3443 (2.1176) -6.3628 (1.91) SC05 2.5816 (1.6026) SC06 0.1875 (0.2351) SC07 0.2078 (0.9482) 1.6853* (0.94) 0.3458 (0.90) -2.0592 (0.8975) -2.5297 (0.92) SC08 -1.0561 (0.4358) SC09 -4.0957 (3.2253) SC10 0.5896 (0.2662) SC11 0.1649 (0.1321) SC12 1.9302 (0.4784) SC13 1.2146 (3.6770) 1.2306 (3.77) -8.7054 (4.01) SC15 0.2816 (0.2482) SC16 0.0928 (0.2017) SC18 0.4750* (0.2617) SC20 -0.5927 (0.2041) SC21 -1.1930 (2.3930) SCtotal 0.1496 (0.1493) Note: 1. * Indicates significance at the 10% level , at the 5% level, and at the 1% level 2. Numbers inside parentheses are the standard deviation of the coefficients 75
38. International science conference “International trade - Policies and practices in Vietnam”, ISBN: 978 - 604 - 67 - 1403 - 3 Table 8: Short-run coefficient estimates of NEG in the NARDL models Product code Lag 0 Lag 1 Lag 2 Lag 3 Lag 4 Lag 5 SC01 1.5806 (1.6938) -0.8967 (1.7433) 2.6599 (1.69) SC02 1.4402 (2.1406) -3.7501* (2.1742) -1.8929 (2.18) -1.67 (1.95) 4.8826 (1.93) -3.86 (1.91) SC04 -0.4087 (1.3354) -3.9021 (1.3486) 2.6566* (1.39) -2.05 (1.37) SC05 9.2575 (5.5768) -6.9226 (5.7963) 1.1087 (5.78) -9.84* (5.59) -15.08 (5.56) SC06 0.3625 (0.8371) -2.6638 (0.8636) 1.3761 (0.89) -2.32 (0.86) SC07 1.0675 (0.6482) -0.9635 (0.6173) SC08 2.1545 (1.6145) 2.0698 (1.6398) 2.9105* (1.64) SC09 1.7635 (2.2227) -2.9568 (1.9745) SC10 0.2238* (0.1146) SC11 -0.74 (0.0820) SC12 -0.4180 (0.2026) SC13 -1.4387 (2.4590) -1.0766 (2.5159) 2.5874 (2.45) SC15 0.1052 (0.8816) 1.5758* (0.8978) -0.7101 (0.91) 0.68 (0.89) 1.6741* (0.87) SC16 0.0559 (0.0979) SC18 0.6133 (0.6985) 0.3194 (0.7099) 1.7874 (0.69) SC20 -0.1574* (0.0884) SC21 1.5308 (0.2696) SCtotal 0.0210 (0.0758) Note: 1. * Indicates significance at the 10% level , at the 5% level, and at the 1% level 2. Numbers inside parentheses are the standard deviation of the coefficients 76