Assessing the relationship between international market and agricultural commodity export prices: evidence from Vietnamese coffee

DALAT UNIVERSITY JOURNAL OF SCIENCE Volume 10, Issue 4, 2020 57-73 57 ASSESSING THE RELATIONSHIP BETWEEN INTERNATIONAL MARKET AND AGRICULTURAL COMMODITY EXPORT PRICES: EVIDENCE FROM VIETNAMESE COFFEE Dang Trung Tuyena, Zhang Caihonga*, Nguyen Thi Honga aBeijing Forestry University, Beijing, P.R.C *Corresponding author: Email: zhangcaihong@263.net Article history Received: February 28th, 2020 Received in revised form (1st): May 26th, 2020 | Received in revised form (2nd): September 19th, 2020 Accepted: September 23rd, 2020 Abstract The paper used a cointegration test, the Granger causality test, and a vector autoregression (VAR) model to determine the relationship between the international coffee price on the spot market and the Vietnamese coffee export price from January 2004 to December 2017. The study found international coffee prices to have a significant effect on Vietnamese coffee export prices, but not vice versa. The two variables are not cointegrated with each other at the 99 percent confidence level, but the Granger causality test confirmed that the Vietnamese coffee export price is influenced by the international market price, while the international market price is not influenced by the Vietnamese coffee export price. The results from the VAR model also showed that the dependent variable is mainly impacted by two independent variables in lag 1 and other lags. Overall, the Vietnamese coffee export price did not have an effect on the international coffee spot market price. Therefore, the relationship between the international coffee price and the Vietnamese coffee export price is asymmetric. These results are in accordance with the actual situation since Vietnam is the largest exporter of robusta coffee in the world, but Vietnam is only a “small” country that has no market power in the international coffee market. Keywords: Co-integration test; Granger causality; VAR model; Vietnamese coffee price. DOI: Article type: (peer-reviewed) Full-length research article Copyright © 2020 The author(s). Licensing: This article is licensed under a CC BY-NC 4.0 DALAT UNIVERSITY JOURNAL OF SCIENCE [ECONOMICS AND MANAGEMENT] 58 ĐÁNH GIÁ MỐI QUAN HỆ GIỮA GIÁ THỊ TRƯỜNG QUỐC TẾ VÀ GIÁ HÀNG NÔNG SẢN XUẤT KHẨU: BẰNG CHỨNG TỪ CÀ PHÊ VIỆT NAM Đặng Trung Tuyếna, Zhang Caihonga*, Nguyễn Thị Hồnga aĐại học Lâm nghiệp Bắc Kinh, Bắc Kinh, Trung Quốc *Tác giả liên hệ: Email: zhangcaihong@263.net Lịch sử bài báo Nhận ngày 28 tháng 02 năm 2020 Chỉnh sửa lần 01 ngày 26 tháng 5 năm 2020 | Chỉnh sửa lần 02 ngày 19 tháng 9 năm 2020 Chấp nhận đăng ngày 23 tháng 9 năm 2020 Tóm tắt Nghiên cứu sử dụng kiểm định đồng liên kết, quan hệ nhân quả Granger và mô hình VAR để xác định mối quan hệ giữa giá cà phê quốc tế trên thị trường giao ngay và giá cà phê Việt Nam xuất khẩu từ tháng 01 năm 2004 đến tháng 12 năm 2017. Nghiên cứu đã tìm thấy ảnh hưởng của giá cà phê thế giới lên giá cà phê Việt Nam xuất khẩu, nhưng không có chiều ngược lại. Hai biến này không có mối quan hệ đồng liên kết ở độ tin cậy 99%, nhưng kiểm định quan hệ nhân quả Granger lại chỉ ra rằng giá cà phê Việt Nam xuất khẩu chịu ảnh hưởng của của giá cà phê trên thị trường thế giới, nhưng giá cà phê trên thị trường thế giới lại không chịu ảnh hưởng bởi giá cà phê Việt Nam xuất khẩu. Những kết quả từ việc hồi quy mô hình VAR cũng chỉ ra rằng, biến phụ thuộc chịu ảnh hưởng của hai biến độc lập ở độ trễ 1 và các độ trễ khác. Tóm lại, giá cà phê Việt Nam xuất khẩu không hề có ảnh hưởng lên giá cà phê quốc tế trên thị trường giao ngay. Do vậy, mối quan hệ giữa hai biến là mối quan hệ phi đối xứng. Những kết quả nghiên cứu này phù hợp với thực tế, mặc dù Việt Nam là quốc gia xuất khẩu cà phê robusta lớn nhất nhưng lại chỉ là một nước “nhỏ” không có bất cứ sức mạnh thị trường nào trên thị trường cà phê thế giới. Từ khóa: Giá cà phê Việt Nam; Kiểm định đồng liên kết; Mô hình VAR; Quan hệ nhân quả Granger. DOI: Loại bài báo: Bài báo nghiên cứu gốc có bình duyệt Bản quyền © 2020 (Các) Tác giả. Cấp phép: Bài báo này được cấp phép theo CC BY-NC 4.0 Dang Trung Tuyen, Zhang Caihong, and Nguyen Thi Hong 59 1. INTRODUCTION Coffee is one of the worldwide most traded commodities, produced in more than 70 developing countries and consumed mainly in developed countries with over US $30.90 billion of total world trade (OEC, 2018). Commercial coffee consists of two main varieties, namely, arabica coffee and robusta coffee. Vietnam is the largest exporter of robusta coffee in the world with an export value of US $3.21 billion (General Statistics Office of Vietnam, 2017). During the fourteen-year period of 2004-2017, the Vietnamese coffee export price moved almost in parallel with the world's robusta coffee price. After a long period of steadily climbing from 30.32 US cents/lb. in 2004 to 101.77 cents/lb. in 2008, Vietnamese coffee export prices had a period of adjustment in the three years from 2008 to early 2011, and the transaction price fell to only 70.00 US cents/lb. The price peaked at 110.16 US cents/lb. in May 2011 before hitting a low of 74.71 US cents/lb. in April 2016. Currently, the Vietnam coffee export price is being traded at approximately 100.00 US cents/lb. (FOB, Ho Chi Minh City price). Despite the fact that the Vietnamese coffee export price is close to the robusta coffee price, it is very difficult to predict since this price fluctuates without any rules (see Figure 1). Figure 1. The fluctuating trend of Vietnamese coffee export price and robusta coffee price Source: General Statistics Office of Vietnam (2017) and the United Nations Conference on Trade and Development (UNCTAD). Against this background, a profound understanding of the relationship between the two markets is of significance to coffee farmers, exporter-producer companies, and the government of Vietnam to accurately forecast future volatility, which is a critical input to the risk management of price volatility. The purpose of this study, therefore, is to determine the relationship between the international market price and the agricultural commodity export price, with a new research subject and from a new perspective. 0 20 40 60 80 100 120 140 2 0 0 4 M 1 2 0 0 4 M 8 2 0 0 5 M 3 2 0 0 5 M 1 0 2 0 0 6 M 5 2 0 0 6 M 1 2 2 0 0 7 M 7 2 0 0 8 M 2 2 0 0 8 M 9 2 0 0 9 M 4 2 0 0 9 M 1 1 2 0 1 0 M 6 2 0 1 1 M 1 2 0 1 1 M 8 2 0 1 2 M 3 2 0 1 2 M 1 0 2 0 1 3 M 5 2 0 1 3 M 1 2 2 0 1 4 M 7 2 0 1 5 M 2 2 0 1 5 M 9 2 0 1 6 M 4 2 0 1 6 M 1 1 2 0 1 7 M 6 PRB-EU PVN DALAT UNIVERSITY JOURNAL OF SCIENCE [ECONOMICS AND MANAGEMENT] 60 Significantly, the price relationship between two commodities or two markets has been studied by many authors. Some papers, such as Acosta, Ihle, and Robles (2014); Alom, Ward, and Hu (2011); Baffes and Gardner (2003); Greb, Jamora, Mengel, Cramon- Taubadel, and Wurriehausen (2016); Hernandez and Torero (2010); Huang, Yang, and Hwang (2009); Minot (2010); Wang and Ke (2005); Zhao and Goodwin (2011) and others used separately several popular models and methods, such as cointegration tests, Granger causality tests, multiple linear regression models, vector autoregressive models (VAR), threshold vector autoregressive (TVAR), autoregressive conditional heteroscedasticity (ARCH) family, and generalized autoregressive conditional heteroscedasticity (GARCH) family models to analyze price transmission comprehensively. In addition, a few studies have addressed price volatility and used combined models (VAR/VECM-DCC-GARCH, VAR/VECM-BEKK-GARCH model), such as Ceballos, Hernandez, Minot, and Robles (2017); Hernandez, Ibarra, and Trupkin (2013); Lee and Valera (2016); Rapsomanikis and Mugera (2011) to deeply analyze the price transmission and volatility spillovers. The main subjects of these studies; however, are the main agricultural commodities, including corn, soybeans, milk, grain, and wheat. In addition, the outcomes of these studies have had mixed results. There are also some researches by Vietnamese scholars on coffee prices in general and Vietnamese coffee prices in particular. These studies mainly used multiple linear regression models and cointegration tests (To, 2015, 2016; Nguyen & Tran, 2015). However, the biggest drawback of multiple linear regression models when estimating time series data is spurious regression if the variables are not stationary at level (Ferson, Sarkissian, & Simin, 2003; Granger & Newbold, 1974; Hamilton, 1994; McCallum, 2010). To the best of our knowledge, the linkages between the world’s robusta coffee price in the spot market and the Vietnamese coffee export price have not been studied by applying a VAR model. In order to address the limitations of previous studies, therefore, this study uses a cointegration test, the Granger causality test, and a VAR model to identify this relationship as well as its trend, equation, and level. 2. LITERATURE REVIEW As explained above, an enormous amount of literature has used various models, including multiple linear regression models, VAR models, VEC models, ARCH-family and GARCH-family models to find linkages among time series variables. The relationship among time series price variables, including the price transmission between the futures market and spot markets, between international markets and domestic markets, and among commodities has been studied by many researchers, including Baffes and Gardner (2003); Ceballos et al. (2017); Gemech and Struthers (2007); Hong (2015, 2016); Lee and Valera (2016); Phuc and Hong (2014); Rahayu, Chang, and Anindita (2015); Worako, Jordaan, and Van Schalkwyk (2011) among others. Most of these studies concerned the price transmission or the long-run and short-run equilibrium relationships among markets or commodities. Dang Trung Tuyen, Zhang Caihong, and Nguyen Thi Hong 61 A few studies used VAR, ECM, or VEC models to test the level of transmission of the world price into the domestic and international prices of commodities. For example, Baffes and Gardner (2003) used a VEC model to study eight countries and ten commodities. They found that only Mexican, Chilean, and Argentinean prices allowed a significant pass-through of world price movements. The transmission between the prices of other countries was found to be either low or nonexistent, and the variability of world prices was not reflected in domestic price movements in any significant way. In 2010, Minot used an ECM model to determine the degree of price transmission between world food market prices and the prices of staple foods in sub-Saharan Africa based on more than 60 price series from 11 African countries. The results indicated a long-term relationship with world prices in only 13 out of the 62 African food prices examined. African rice prices are more closely linked to world markets than corn prices. Acosta et al. (2014) also used an ECM model to examine the price transmission from the global market to the domestic market in Panama. They showed that there is a long-run cointegration relationship between global and domestic producers’ prices; however, only producers' prices showed significant responses to price disequilibria, which appears to be plausible due to the relative sizes of both markets. Furthermore, the results of the ECM pointed out the potential presence of asymmetric price transmission of global and domestic milk prices, indicating that increases in global prices tend to be transmitted faster to producers’ price than decreases. For forecasting commodity prices of rice and coffee in Vietnam, Nguyen and Tran (2015) developed an efficient maximum likelihood estimation procedure based on the characteristics function. They estimated parameters of a stochastic volatility model with stochastic drift utilizing the time series. Finally, by using the estimated model parameters, they calculated various risk measures, such as value at risk and expected shortfall. Hong (2015) used multiple linear regression models to identify and measure the impact of such factors as exchange rates and gasoline prices on the price of Vietnamese coffee exports for of 2008-2014. A cointegration between the Vietnamese coffee price and the world coffee price, was found. Based on the pairwise Granger test, Hong (2015) found that the world coffee price could affect the Vietnamese coffee price, but not vice-versa. The very significant result of this study is that exchange rates and gasoline prices can affect the export price of Vietnamese coffee. In 2016, Hong again found that Vietnamese coffee prices vary with the trends in world coffee prices. The author used data on the price of Vietnamese coffee over 34 years with cyclic changes consisting of a five-year increase followed by a seven-year decrease. The Brazilian coffee price was found to have an important effect on the Vietnamese coffee export price, which was estimated to go up by 0.31% with a 1.00% increase in the Brazilian coffee price. There is already a large empirical literature on the relationship between international market prices and domestic prices. However, the conclusions of the literature appear to be mixed. The differences probably arise from different time periods, datasets, frequencies, methodologies, utilized models, and kinds of commodities. Furthermore, although there is some evidence of the influence of these factors on coffee export price, it is still not clear and detailed, especially in the literature on Vietnamese coffee. DALAT UNIVERSITY JOURNAL OF SCIENCE [ECONOMICS AND MANAGEMENT] 62 In an effort to overcome the limitations of previous research and the lack of multiple regression models, this paper attempts to analyze the linkage between the international coffee price on the spot market and a new research subject–the Vietnamese coffee export price–by applying a cointegration test, the Granger causality test, and a VAR model. 3. RESEARCH METHODOLOGY 3.1. Co-integration test In the 1980s, Engle and Granger proposed the concept of cointegration (Engle & Granger, 1987). If the time series (t = 1, 2, ...) becomes stationary after d differences, and the sequence difference is d – 1 times, then the sequence Xt is called a d-ordered single integer sequence, denoted as Xt ~ I(d). If two time series Xt and Yt are both I(d), any linear combination of Xt and Yt will be also I(d). If, however, there exists a vector, such that the combination st= aXt + bYt is I(d-c) (d ≥ c≥ 0), then Xt and Yt are called (d-c) order cointegrated. For those time series variables that are non-stationary, if some of their linear combinations are stationary, the linear combination reflects the long-term equilibrium relationship between the variables, which is the cointegration relationship. Testing cointegration is a significant step to check whether empirically meaningful relationships exist in the model or not. If variables have different trend processes, they cannot stay in a fixed long-run relationship, implying that you cannot model the long-run, and there is usually no valid basis for inferences based on standard distributions. If cointegration is not found, it is necessary to continue working with variables in differences instead. The cointegration relationship among variables can be tested by using the Johansen cointegration method (Johansen & Juselius, 1990) and the Engle-Granger two-step cointegration method (Engle & Granger, 1987). 3.2. Granger causality test The Granger causality test (Granger, 1969) assumes that most of the information about the predictions of y and x is contained in their time series. The test requires the estimation of y and x by the following regressions: 1 1 1 q q t i t i j t j t i j y x y u − − = = = + +  (1) 2 1 1 s s t i t i j t j t i j x x y u − − = = = + +  (2) where xt, yt, represent two variables; yt-j, xt-i denote the lag of yt, xt, respectively; αi, βj, λi, δj denote the coefficient estimation of the lag term; i, j, q, s denote lag order; u1t and u2t are white noise and assumed to be irrelevant. Dang Trung Tuyen, Zhang Caihong, and Nguyen Thi Hong 63 Equation (1) assumes that the current y relates to lags of y itself and past values of x, and Equation (2) assumes similar behavior for x. For Equation (1), the null hypothesis is H0: α1 = α2 = = αq=0; for Equation (2), the null hypothesis is H0: δ1 = δ2 = = δs = 0. Values of the F-statistic and p-value will be used to decide to not reject the null hypothesis if the p-value is greater than 5% or to reject it if the p-value is less than 5%. 3.3. VAR model and VEC model The vector autoregression model (VAR) was introduced as a technique that can be used by macroeconomists to characterize the joint dynamic behavior of a collection of variables without requiring strong restrictions to identify underlying structural parameters (Sims, 1980). It has become a prevalent method of time series modeling. The VAR model can be expressed as: zt = A1zt-1 + A2zt-2 + + Apzt-p + Bvt + εt (3) where zt is a k-dimensional vector of the endogenous variables, t is the number of observations, p is the order of the lagged variable, and vt is the d-dimensional exogenous variable vector. The (k x k) dimensional matrices A1, ..., Ap and (k × d) dimensional matrix B are the coefficient matrices to be used for estimation, and εt is a vector of k-dimensional disturbances. Generally, p and R-squared must be large enough to fully reflect the dynamic characteristics of the VAR model. But the accuracy of the model does not depend on how big p is, so an equilibrium must be established between p and R-squared. This equilibrium can be determined by five criteria: LR, FPE, AIC, SC, and HQ. The Eviews 8.0 software was used to estimate and test the hypotheses of the above model. 4. EMPIRICAL ANALYSIS 4.1. Selecting variables, model and data description 4.1.1. Selecting variables and data description Because all of the Vietnamese coffee traded on the spot market is robusta coffee, the robusta coffee price on the spot market directly and immediately influences Vietnam’s future domestic coffee price. Nowadays, robusta coffee is mainly traded on the London International Financial Futures and Options Exchange (LIFE), as are many primary commodities such as rice, gold, copper, oil, coffee, sugar, and so on. These commodities are traded on two markets, the spot market and the futures market. For the reasons given above, this study selected two variables, the Vietnamese coffee export price and the robusta coffee price on the spot market (representing the international price). Their price histories are shown in Figure 1. The monthly average price of the above two variables is based on data collected from January 2004 to December 2017, with 168 observations for each variable. The DALAT UNIVERSITY JOURNAL OF SCIENCE [ECONOMICS AND MANAGEMENT] 64 Vietnamese coffee export price (PVN) was converted from the Government Statistics Office of Vietnam reports, and the robusta coffee price in the spot market trading on LIFE (PRB_EU) was converted from UNCTAD. All of the data have been seasonally adjusted using Census X12. Descriptive statistics of the variables can be seen in Table 1. Table 1. Descriptive statistics Statistical indicators PVN PRB_EU Mean 78.1624 83.6566 Median 84.2567 88.4101 Maximum 108.4201 122.9824 Minimum 28.0353 33.8568 Std. Dev. 23.4345 21.6552 Skewness -0.7705 -0.7917 Kurtosis 2.4319 2.8144 Jarque-Bera 18.8796 17.7920 Probability 0.0001 0.0001 Sum 3131.2800 14054.3000 Sum Sq. Dev. 91712.1600 78314.5400 Observations 168.0000 168.0000 Source: Calculated by the authors using Eview 8. As can be seen in Table 1, the mean value of the Vietnamese coffee export price is less than that of the international robusta coffee price, while the value of the standard deviation of the Vietnamese coffee export price is not. Hence, the Vietnamese coffee export price is less than the international robusta coffee price, but its fluctuation is greater. The two variables are skewed left (skewness is less than 0). In addition, the values of the Jarque-Bera test and the p-values indicate that the two variables are not normally distributed. 4.1.2. Selecting the model • Unit root test Table 2. ADF and Phillips Perron test results Variable (C,T,L) ADF t-statistic PP t-statistic 1% level 5% level Conclusions PVN (1,1,13) -1.8489 -2.0156 -4.0143 -3.4371 Non-stationary DPVN (1,1,13) -9.6096 -9.7825 -4.0143 -3.4371 Stationary PRB_EU (1,1,13) -2.1702 -2.1209 -4.0143 -3.4371 Non-stationary DPRB_EU (1,1,13) -9.7707 -9.7703 -4.0143 -3.4371 Stationary Notes: C is constant or intercept, T is trend, and L is lag selection; D represents the first-order difference of the time series. Source: Calculated by the authors using Eviews. Dang Trung Tuyen, Zhang Caihong, and Nguyen Thi Hong 65 The estimation will start with the unit root test. According to Schwert (2002) the lag difference is (Pmax) = [12.(T/100)1/4] where T is the number of observations. The model uses monthly data from January 2004 to December 2017 with 168 observations. Therefore, Pmax=13. Thirteen lags are used to test that the variables are stationary. The stationarity is tested using the Augmented Dickey-Fuller (ADF) and Phillips Perron tests (PP). Table 2 describes the ADF and PP test results at the level and first difference. In both tests, the null hypothesis is that there exists a unit root for each variable. After the test, the two variables are stationary at first difference at the 1% level, meaning that all of the variables are integrated in the same order. • Determination of Lags Table 3 reports lag-order selection statistics. The authors use the lowest value of five criteria, namely, LR (Likelihood - Ratio), FPE (Final Prediction Error), AIC (Akaike Information Criterion), SC (Schwarz Information Criterion), and HQ (Hannan-Quinn Information Criterion) as a primary concern to determine the lag length (Ng & Perron, 2001). Based on the result of these criteria (which is indicated by an asterisk in Table 3), we perform further tests with Lag (4). Table 3. Determining Lag Length for the VAR Model Lag LR FPE AIC SC HQ 0 NA 16466.4800 15.3848 15.4233 15.4005 1 816.7044 95.3116 10.2329 10.3482 10.2797 2 34.8281 80.0368 10.0582 10.2504* 10.1362* 3 10.5444 78.5425 10.0393 10.3084 10.1486 4 13.6186* 75.4580* 9.9991* 10.3451 10.1396 5 2.2930 78.1308 10.0337 10.4565 10.2054 6 7.7903 77.9192 10.0307 10.5304 10.2336 7 1.5810 81.0572 10.0698 10.6464 10.3039 8 1.2971 84.4864 10.1107 10.7642 10.3761 Source: Calculated by the authors using Eviews. 4.2. The empirical analysis 4.2.1. Co-integration test Cointegration rank is estimated using Johansen methodology. Johansen’s approach (Johansen & Juselius, 1990) derives from two likelihood estimators for the cointegration rank: a trace test and a maximum eigenvalue test. The results of the cointegration test are shown in Table 4. The results show that there is no cointegrating equation at the 0.05 level between the variables, meaning there is no long-term relationship between the two variables from 2004 to 2017. This study does not find the DALAT UNIVERSITY JOURNAL OF SCIENCE [ECONOMICS AND MANAGEMENT] 66 same results as Hong (2016) and Phuc and Hong (2014). The differences may come from the difference in selecting the number of variables and the time interval. From our results, the VAR (4) model is the most suitable. Table 4. Cointegration rank test Unrestricted Cointegration Rank Test (Trace) Hypothesized Eigenvalue Trace 0.0500 Prob.** No. of CE(s) Statistic Critical Value None 0.0609 14.1086 15.4947 0.0800 At most 1* 0.0235 3.8737 3.8415 0.0490 Trace test indicates no cointegration at the 0.05 level. Unrestricted Cointegration Rank Test (Maximum Eigenvalue) Hypothesized Eigenvalue Max-Eigen. 0.0500 Prob.** No. of CE(s) Statistic Critical Value None 0.0609 10.2349 14.2646 0.1970 At most 1* 0.0235 3.8737 3.8415 0.0490 Notes: Maximum eigenvalue test indicates no cointegration at the 0.05 level. Source: Calculated by the authors using Eviews. 4.2.2. Granger causality test The results of the cointegration test show that there is no long-term equilibrium between the two variables, but only a short-term relationship. Therefore, the causal analysis needs further verification. A Granger causality test with a Lag (4) was conducted to verify the causal relationship between the two variables. Estimation results for Granger causality between the variables are presented in Table 5. The authors use F-statistics and probability to test causality between the variables with the null hypothesis of no Granger cause among those variables. Table 5. Partial Granger causality test results Null Hypothesis Obs. F-Statistic Prob. D(PRB_EU) does not Granger cause D(PVN) 164.0000 18.3100 3.E-12 D(PVN) does not Granger cause D(PRB_EU) 1.0500 0.3834 Sources: Calculated by the authors using Eviews. Table 5 provides the results of the pairwise analysis. Significant probability values denote rejection of the null hypothesis. In this study, based on the F-statistic and p-value, the null hypothesis will be rejected if the p-value is less than 0.05, and vice versa. Then, PRB_EU has “Granger cause” with PVN unidirectional at a 5% significance level, but not vice versa. Hence, Vietnamese coffee absolutely does not have the dominant power in the world coffee market. The relationship found between the robusta coffee spot market price and the Vietnamese coffee export price is very realistic and consistent with the study of Phuc and Hong (2014). Despite the fact that Vietnam is the world's biggest exporter of robusta coffee and the second largest exporter of coffee in the world after Brazil, the price Dang Trung Tuyen, Zhang Caihong, and Nguyen Thi Hong 67 of Vietnamese coffee entirely depends on world coffee prices. The main reason is that more than 90% of the total coffee production is used for export (while this proportion of Brazil is only 60%), and domestic consumption accounts for a small proportion (approximately 10%) (ICO, 2017 and author’s calculation). Being dependent on the world coffee market creates a huge risk for coffee enterprises as well as coffee growers in Vietnam when the price of coffee fluctuates. 4.2.3. Vector autoregression model (VAR model) Because no cointegration leads to no long-run relationship between variables, the VAR model can be applied to analyze the relationship between the two markets in the short-run. The VAR model shows the relationship between the variables from January 2004 to December 2017 with standard errors in parentheses, t-statistics in brackets as follows: Table 6. Unrestricted Vector Autoregression Estimates of PVN and PRB_EU D(PVN) D(PRB_EU) D(PVN(-1)) -0.3447 (0.1024) [-3.3651] -0.1677 (0.1530) [-1.0961] D(PRB_EU(-1)) 0.5634 (0.0678) [8.3042] 0.3500 (0.1014) [3.4535] D(PVN(-2)) -0.2539 (0.1013) [-2.5075] 0.0613 (0.1513) [0.4052] D(PRB_EU(-2)) 0.2662 (0.0804) [3.3125] 0.0017 (0.1201) [0.0142] D(PVN(-3)) 0.0147 (0.1007) [0.1462] -0.2408 (0.1505) [-1.5999] D(PRB_EU(-3)) 0.2011 (0.082) [2.4537] 0.1420 (0.1224) [1.1600] D(PVN(-4)) 0.0222 (0.0841) [0.2640] 0.0159 (0.1257) [0.1269] D(PRB_EU(-4)) 0.0726 (0.0801) [0.9076] 0.1351 (0.1196) [1.1294] C 0.2234 (0.2150) [1.0394] 0.2361 (0.3211) [0.7352] Source: Calculated by the authors using Eview 8. DALAT UNIVERSITY JOURNAL OF SCIENCE [ECONOMICS AND MANAGEMENT] 68 Table 6 shows that the most recent previous price has the largest influence on the current Vietnamese coffee export price. In particular, the price of robusta coffee on the spot market in the most recent previous period has the greatest impact on the current Vietnamese coffee export price. In terms of the direction of influence, among previous periods of Vietnamese coffee export price, the first and the second previous period variables have a negative effect on the dependent variable, while the others have a positive impact on PVN. Meanwhile, all previous periods of robusta coffee spot market price have a positive impact on PVN. 4.2.4. Impulse response function (IRF) and variance decomposition analysis In the final step of the empirical modeling analysis, the author defines the response to PVN when there is a shock in the international market price and itself. In this regard, the generalized impulse-response functions derive from the VAR model for two variables. The optimal lag lengths in the VAR system are determined via the Schwartz information criterion with Lag (4). -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Inverse Roots of AR Characteristic Polynomial Figure 2. Inverse roots of AR characteristic polynomial Source: Drawn by the authors using Eviews. Figure 2 shows that all roots of the VAR model are within the unit circle. Hence, the VAR system satisfies the stability condition. • Impulse Response Function (IRF) In practical applications, the VAR model generally does not analyze how the change of one variable affects another variable. It examines the dynamic structural analysis of variables as well as the dynamic influence of one error term of the model or the overall impact of receiving some kind of shock. Also, the economic interpretation of the single parameter estimation is relatively difficult when the impulse response analysis is required. This article Dang Trung Tuyen, Zhang Caihong, and Nguyen Thi Hong 69 selects the most commonly used analysis, Cholesky orthogonal impulse response, which is shown in Figure 3. -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 1 2 3 4 5 6 7 8 9 10 Response of D(PVN_SA) to D(PRB_EU_SA) -1 0 1 2 3 4 1 2 3 4 5 6 7 8 9 10 Response of D(PRB_EU_SA) to D(PVN_SA) Figure 3. Response to Cholesky one S.D. Innovation +/- 2SE Source: Drawn by the authors using Eviews. This section analyzes the response of the robusta coffee price to the Vietnamese coffee export price and vice versa. Figure 3 illustrates that when getting a shock, the changes in Vietnamese coffee export price will directly affect the international market price, but this response is not stable. The maximum value is approximately 1.7% in period 1 and it decreases to zero after nine periods. On the other hand, the response of the international market price to a shock in the Vietnamese coffee export price is not steady either. The trend is mainly decreasing from period 1 to period 10. Overall, the relationship between the two price series is closely related since the EU is the biggest market for Vietnamese coffee exports. The influence of the international market price on the Vietnamese coffee export price is higher than the impact of the Vietnamese coffee export price on the international market price. After determining the response of the variables to a shock, this study next examines the volatility variance of the Vietnamese coffee export price under the influence of the international market price. • Variance decomposition analysis From Table 7, when the forecast period is 1, the variance of the Vietnamese coffee export price changes from 100%, while the contribution from the international market price is zero percent. There is a slight decline from 69.6% in period 2 to 64.5% in period 10. The contribution from the international market price to the change in the variance of the Vietnamese coffee export price is not in the same, it increases from zero in period 1 to 35.5% in period 10. The study also observes 36.7% changes in the international market price from the level of the Vietnamese coffee export price, which is 63.3% of their participation in period 1. The analysis found that as the forecast period increased, the effect of the international market price to self-variance is not stable, while the contribution of the Vietnamese coffee export price to the variance of international market price is the same. At the tenth forecast period, the variation of the international market price data is 65.4% from itself and 34.6% from the Vietnamese coffee export price. DALAT UNIVERSITY JOURNAL OF SCIENCE [ECONOMICS AND MANAGEMENT] 70 Table 7. Variance decomposition of PVN and PRB_EU Variance decomposition of PVN Variance decomposition of PRB_EU Period S.E. D(PVN_SA) D(PRB_EU_SA) S.E. D(PVN_SA) D(PRB_EU_SA) 1 2.6668 100.0000 0.0000 3.9839 36.6784 63.3216 2 3.2396 69.6076 30.3924 4.1546 34.6409 65.3591 3 3.3503 65.0956 34.9045 4.1622 34.8277 65.1723 4 3.4186 65.9906 34.0094 4.1912 34.5668 65.4332 5 3.4497 64.8208 35.1792 4.2064 34.4376 65.5624 6 3.4584 64.5283 35.4717 4.2099 34.5383 65.4617 7 3.461 64.5739 35.4261 4.2108 34.5580 65.4420 8 3.4617 64.5473 35.4527 4.2112 34.5697 65.4303 9 3.4621 64.546 35.454 4.2115 34.5775 65.4225 10 3.4622 64.5469 35.4531 4.2115 34.5775 65.4225 Source: Calculated by the authors using Eviews. Overall, from the order of the variance decomposition analysis, the response of the international market price to the Vietnamese coffee export price is slightly higher than the reaction of the Vietnamese coffee export price to the international coffee price. This evidence also illustrates that there is a relationship between the two price series, of which the international market price plays a guiding role. 5. CONCLUSIONS The fluctuation of agricultural product prices is always a difficult issue for countries with developing agricultural sectors, like Vietnam. Coffee export prices, for example, always fluctuate unpredictably, which directly affects not only the export turnover but a large number of farmers’ incomes as well. Using time series data from January 2004 to December 2017, this paper investigated the price transmission of the international market price on the Vietnamese coffee export price. The results demonstrated that the two variables are not cointegrated with each other at the 99% confidence level. The Granger causality test confirmed that the Vietnamese coffee export price is influenced by the international market price, but not vice versa. The results from this study also showed that the dependent variable is mainly impacted in period 1 by two independent variables with coefficients of 0.3447 and 0.5634, respectively. Meanwhile, the Vietnamese coffee export price did not affect the robusta coffee spot market price. Therefore, the transmission from the international coffee price to the Vietnamese coffee export price is asymmetric. The Vietnamese coffee export price did not have an influence on the world coffee spot market price. Vietnam is only a “small” country in the international coffee market, although Vietnam is the second biggest exporter and producer of robusta coffee in the world. The world coffee price provides a basis to calculate the Vietnamese coffee export price, but not vice versa. This conclusion is in accordance with the actual situation of the Vietnam coffee industry nowadays. Dang Trung Tuyen, Zhang Caihong, and Nguyen Thi Hong 71 In summary, the findings of this article contribute in a practical way to further improving studies of price transmission using time series data. The findings provide further evidence confirming the relationship between international market prices and agricultural commodity export prices. In regard to practical significance, the results of this study can help coffee enterprises and coffee farmers clearly understand the fluctuation of Vietnamese coffee export price in order to form appropriate strategies. However, there are still some limitations in this paper. The restrictions can be a hint for future research. In that respect, future studies can extend the literature at least in two ways so as to provide some new insights. Firstly, a VAR model could still be applied, but the number of independent variables or the time interval could be extended. Secondly, researchers could increase the number of independent variables and use another model, such as a structure vector autoregressive model, a panel vector autoregressive model, a Markov switching vector autoregressive, or combine vector autoregressive- and GARCH- family models with the same data. These approaches would probably provide significant new research insights into the relationship between international market prices and agricultural product export prices. REFERENCES Acosta, A., Ihle, R., & Robles, M. (2014). Spatial price transmission of soaring milk prices from global to domestic markets. Agribusiness, 30(1), 64-73. Alom, F., Ward, B. D., & Hu, B. (2011). Spillover effects of World oil prices on food prices: Evidence for Asia and Pacific countries. Paper presented at The Proceedings of the 52nd Annual Conference New Zealand Association of Economists, Wellington, New Zealand. Baffes, J., & Gardner, B. (2003). The transmission of world commodity prices to domestic markets under policy reforms in developing countries. The Journal of Policy Reform, 6(3), 159-180. Ceballos, F., Hernandez, M. A., Minot, N., & Robles, M. (2017). Grain price and volatility transmission from international to domestic markets in developing countries. World Development, 94, 305-320. Engle, R. F., & Granger, C. W. (1987). Co-integration and error correction: Representation, estimation, and testing. Econometrica, 55(2), 251-276. Ferson, W. E., Sarkissian, S., & Simin, T. T. (2003). Spurious regressions in financial economics? The Journal of Finance, 58(4), 1393-1413. Gemech, F., & Struthers, J. (2007). Coffee price volatility in Ethiopia: Effects of market reform programmes. Journal of International Development, 19(8), 1131-1142. General Statistics Office of Vietnam. (2017). Social and economic situation in 2017. Retrieved from https://www.gso.gov.vn/default_en.aspx?tabid=622&ItemID=18670 Granger, C. W. J. (1969). Investigating causal relations by econometric models and cross- spectral methods. Econometrica, 37(3), 424-438. DALAT UNIVERSITY JOURNAL OF SCIENCE [ECONOMICS AND MANAGEMENT] 72 Granger, C. W. J., & Newbold, P. (1974). Spurious regressions in econometrics. Journal of Econometrics, 2(2), 111-120. Greb, F., Jamora, N., Mengel, C., Cramon-Taubadel, V., & Wurriehausen, N. (2016). Price transmission from international to domestic markets. Washington, USA: World Bank Group Publishing. Hamilton, J. D. (1994). Time series analysis (Vol. 2). New York, USA: Princeton University Press Publishing. Hernandez, M., & Torero, M. (2010). Examining the dynamic relationship between spot and future prices of agricultural commodities (IFPRI Discussion Paper No. 00988). Retrieved from https://www.ifpri.org/publication/examining-dynamic- relationship-between-spot-and-future-prices-agricultural-commodities. Hernandez, M. A., Ibarra, R., & Trupkin, D. R. (2013). How far do shocks move across borders? Examining volatility transmission in major agricultural futures markets. European Review of Agricultural Economics, 41(2), 301-325. Hong, T. T. K. (2015). Effect of exchange rates and gasoline price on export price of Vietnamese coffee. Journal of Science Ho Chi Minh City Open University, 4(16), 29-36. Hong, T. T. K. (2016). The volatility and competitiveness of Vietnam coffee export on world market. Van Hien University Journal of Science, 4(3), 85-92. Huang, B.-N., Yang, C. W., & Hwang, M. J. (2009). The dynamics of a nonlinear relationship between crude oil spot and futures prices: A multivariate threshold regression approach. Energy Economics, 31(1), 91-98. ICO. (2017). Record Exports for Coffee Year 2016/17. Retrieved from org/show_news.asp?id=635. Johansen, S., & Juselius, K. (1990). Maximum likelihood estimation and inference on cointegration–with applications to the demand for money. Oxford Bulletin of Economics and Statistics, 52(2), 169-210. Lee, J., & Valera, H. G. A. (2016). Price transmission and volatility spillovers in Asian rice markets: Evidence from MGARCH and panel GARCH models. The International Trade Journal, 30(1), 14-32. McCallum, B. T. (2010). Is the spurious regression problem spurious? Economics Letters, 107(3), 321-323. Minot, N. (2010). Transmission of world food price changes to markets in Sub-Saharan Africa (IFPRI Discussion Paper No. 01059). Retrieved from https://www.ifpri. org/publication/transmission-world-food-price-changes-market s-sub-saharan-africa. (2011). Ng, S., & Perron, P. (2001). Lag length selection and the construction of unit root tests with good size and power. Econometrica, 69(6), 1519-1554. Nguyen, T. N. T., & Tran, N. T. (2015). A methodology to forecast commodity prices in Vietnam. International Journal of Economics and Finance, 7(5), 44-49. Dang Trung Tuyen, Zhang Caihong, and Nguyen Thi Hong 73 OEC. (2018). Retrieved from https://oec.world/en/profile/hs92/20901. Phuc, N. V., & Hong, T. T. K. (2014). Cointegration test for Vietnam’s coffee export price and world coffee price over the period 2008-2014. Journal of Science Ho Chi Minh City Open University, 4(37), 30-36. Rahayu, M. F., Chang, W.-I., & Anindita, R. (2015). Volatility analysis and volatility spillover analysis of Indonesia's coffee price using Arch/Garch, and Egarch model. Journal of Agricultural Studies, 3(2), 37-48. Rapsomanikis, G., & Mugera, H. (2011). Price transmission and volatility spillovers in food markets of developing countries. In I. Piot-Lepetit & R. M'Braek (Eds), Methods to analyse agricultural commodity price volatility (pp. 165-179). London, UK: Springer Publishing. Schwert, G. W. (2002). Tests for unit roots: A Monte Carlo investigation. Journal of Business and Economic Statistics, 7(2), 147-159. Sims, C. A. (1980). Macroeconomics and reality. Econometrica, 48(1), 1-48. To, T. K. H. (2015). Effect of exchange rates and gasoline price on export price of vietnamese coffee. Journal of Science Ho Chi Minh City Open University, 4(16), 29-36. To, T. K. H. (2016). The volatility and competitiveness of Vietnam coffee export on world market. Van Hien University Journal of Science, 4(3), 85-92. Wang, H. H., & Ke, B. (2005). Efficiency tests of agricultural commodity futures markets in China. Australian Journal of Agricultural and Resource Economics, 49(2), 125-141. Worako, T., Jordaan, H., & Van Schalkwyk, H. D. (2011). Investigating volatility in coffee prices along the Ethiopian coffee value chain. Agrekon, 50(3), 90-108. Zhao, J., & Goodwin, B. K. (2011). Volatility spillovers in agricultural commodity markets: An application involving implied volatilities from options markets. Paper presented at The 2011 Annual Meeting, Pennsylvania, USA.