CHAPTER 3: PERFORMANACE OF TRADE BALANCE AND
3.3. The fluctuation of exchange rate in 2000-2010 period
3.3.2. Movement of real exchange rates
3.3.2.2. Analysis bilateral real exchange rate and real effective
* Bilateral real exchange rate
The movement of bilateral real exchange rate between VND and USD can be divided into two phases. From 2000 to end 2004, RER index is higher than 100, showing depreciation of VND against USD in real value. However, the upward trend from 2000 disappears in 2004 and from 2005 RER turns to be overvalued (RER is lower than 100), and keeps being overvalued at higher level till the end of 2010.
Inadequate adjustment between nominal exchange rate change and inflation
45
difference of two countries is the reason standing behind this movement. From 2000 to 2003, that year-on-year change of nominal exchange rate is little and inflation difference of Vietnam and US is inappreciable makes RER gradually rise. While inflation difference turns to be considerable from 2004, nominal exchange rate change keeps remain little. VND, as a result, reverses its trend toward appreciate against USD. After two year of high inflation, VND take back its equilibrium with USD and turn to be highly overvalued for the rest of period when inflation difference between two countries becomes greater.
Figure 3.6 – RER and trade balance
0.00 20.00 40.00 60.00 80.00 100.00 120.00
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 0.000 0.200 0.400 0.600 0.800 1.000 1.200
RER (Left axis) X/M (Right axis)
Source: Trade balance: GSO, RER: calculated by author
It can be seen in Figure 3.6 that the movement of VND’s RER cannot explain for the performance of trade balance (also represented by X/M) because this index indicates the competitiveness between goods of Vietnam and US (due to calculation based on nominal exchange rate between VND and USD and CPI of Vietnam and US) but the general competitiveness of Vietnam’s goods. However, RER’s movement can play an important part to deteriorate Vietnam’s trade balance. According to a research of ADB by Emma Xiaoquin Fan (2001), for country which pegs its currencies to USD, its merchandises competitiveness will be influenced by the fluctuation of USD value. When USD depreciates, by depreciating in line with USD, its products will become more competitive against third country products, then its trade balances with countries other than US are expected to improve. Adversely, when USD appreciates, by appreciating in line with USD, its product will become less competitive and its trade balances with countries other than US are expected to deteriorate. Since 2003,
46
USD has appreciated considerably. The nominal effective exchange rate (NEER) and REER of USD have appreciated by 17 percent and 15 percent respectively between 2003 and 2010 (IFS). In the meantime, the VND’s RER turned to be overvalued since 2005. It causes VND to appreciate much greater against other currencies. Also, as shown earlier in table 3.5, many trading partner with US and Vietnam devaluates or remain unchanged their currencies5. Those may contribute to fast deficit of trade balance from 2005.
* Real effective exchange rate
The shape of VND’REER is a combination of shape of bilateral real exchange rate between VND and USD and USD’s REER (Figure 3.7), once more time showing VND is pegged to USD. This combination especially fits in period 2000-2005 when VND is kept in a narrow range with USD (Table 3.6) and inflation difference is not too large.
In this situation, the long-run movement of VND’REER tends to move downward for the whole period, showing VND is increasing its real value to currency basket and Vietnam’s good is loosing competitiveness. The trade balance with major trading partners, therefore, becomes more and more deficit (Figure 3.8).
Figure 3.7 – VND/USD RER, VND’s REER and USD’s REER.
0.00 20.00 40.00 60.00 80.00 100.00 120.00
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 RER VND's REER USD's REER
Source: RER, VND’s REER: calculated by author; USD’s REER: IFS
5 Some main trading partners of US is also main trading partners of Vietnam, such as: China, Japan, Germany, South Korea, Taiwan, Singapore. Source: US census Bureau and GSO
47
Figure 3.8 – REER and trade balance with major trading partners
0 20 40 60 80 100 120
2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 0 0.2 0.4 0.6 0.8 1 1.2
REER (left axis) X/M (right axis)
Source: REER: calculated by author, X/M: calculated by author from GSO data
However VND’s REER is overvalued to the base year for most of the time during the period and only turns into undervalued from 2007. As in theory, the trade balance of Vietnam must be improved as competitiveness of Vietnam’s good increases.
Meanwhile, the trade balance with major trading partners in reality deteriorates. This performance implies that trade balance not only determined by exchange rate but other macro-economic factors. The impact of other variables must be significant because their impacts on trade balance are so strong that overshadowing exchange rate’s impact, causing trade balance to larger deficit.
Despite the value of REER to the base year cannot explain for the trade balance performance, its movements in trend can do that. From 2000 to 2003, REER moves upwards, the trade balance, also represented by export-to-import ratio (X/M) move downwards. When REER reaches peak in 2003 and decreases in two following years, X/M increases. When REER slightly rises in 2006, X/M continues to slightly increase. And when REER turns back to its downward trend in the next years, X/M decreases in response. It seems that trade balance response is always late to the impact of REER which is involved in J-curve effect theory (Figure 3.8).
In short, that VND is pegged to USD in this period causes value of VND to fluctuate in line with that of USD against other currencies. As real exchange rate of VND against USD appreciates and REER of USD appreciates, REER of VND as a
48
combination of them appreciates significantly, leading to large deficit of Vietnam’s trade balance. Although USD plays an important role to determine the nominal value of VND (VND is keep to be stable with USD), the fluctuation of VND against USD in both nominal and real terms cannot explain for the performance of trade balance.
The movement in trend of REER to some extent can do explain. However, the impact of REER seems to be not as great as impacts of other macroeconomic variables, that makes trade balance to response little to REER fluctuation.
Chapter summary
This chapter has shown the performance of Vietnam’s trade balance which has long-lasting deficit status for the studied period. Despite the deficit of trade balance is understandable for developing country case with large proportion of materials for productions, it large amount is not normal and should not last any longer. In the meantime, as SBV keeps pegging VND to USD while USD appreciates to other currencies, VND appreciates against currencies of other countries who are also main trading partners of Vietnam. The exchange rate movement of Vietnam exposes to have some responsibility for the problem of trade balance.
49
Chapter 4
ESTIMATING THE IMPACT OF REAL EXCHANGE RATE ON TRADE BALANCE IN VIETNAM IN 2000-2010
4.1. Model specification
The study use model (2.5) developed by Tihomir Stucka (2004) based on the standard ―two-country‖ imperfect substitutes model as specified in Goldstein and Kahn (1985) and Rose and Yellen (1989). When applying the theoretical model (2.5) into specific country case, it’s rationale to re-write the following model to be estimated:
TB = β1 + β2REER + β3GDP + β4GDP* + ε (4.1)
In which TB represents trade balance, REER represents real effective exchange rate, GDP represents domestic output and GDP* represents foreign output. The reason for choosing REER as representative for exchange rate is explained earlier. Because a country trades with multiple of partners, only REER can reflect country’s currency value relative to the other currencies, as adjusted for the effect of inflation. GDP and GDP* are chosen because they are the best choice for variables measuring countries’
income.
Following a numbers of previous studies like Bahmani- Oskooee (2001), Tihomir Stučka (2004), Jarita Duasa (2007), Khan and Hossain (2010), Pavle Petrović and Mirjana Gligorić (2010), model 4.1 is in log-linear form. The attractive feature of the log-linear model is that the slope coefficient measures the elasticity of the dependent variable with respect to the independent variable (Khan and Hossain, 2010).
In this model, REER and TB are expected to be positive relation ( >0), indicating that depreciation of currency will improve trade balance, and the other way round in case of depreciation. GDP* and TB are expected to be positive related ( >0) under the rationale that when there is a rise in foreign income, demand for export will increase. The impact of GDP on TB is ambiguous because, an increase in domestic output raises imports but could also boost exports, and the net effect on the trade
50
balance could either be an improvement or a worsening. It is well understood that the supply drives output growth, e.g. due to an increase in productivity, leads to an improvement of the trade balance (Caves, Frankel, and Jones, 2001). Historic examples are those of Germany and Japan in the 1960s and the 1970ss, as well China in the 1990s and the 2000s. On the other hand, the demand driven increase in output, as in e.g. US in the 1970s and the 2000s, end up with trade balance deterioration.
4.2. Data description
4.2.1. Technical data description
In this study, we use quarterly data from 2000(1) to 2010(4) to estimate the impact of exchange rate on trade balance. The choice of this period relies on the availability of data. In detail, quarterly GDP data of Vietnam is not available before 2000.
Besides, as mentioned earlier, one year before 2000, SBV announces to change exchange rate regime of Vietnam to managed floating exchange rate and re-adjusts the nominal value of VND to bring it back to its real value relative to other currencies.
Also, the trade balance of Vietnam is fairly balance in 2000.
REER data is mentioned earlier when explaining REER calculation.
Real GDP data for seventeen (17) trading partners from 2000(1) to 2008(4) is obtained from IFS in index form. For the rest period, 2009(1) to 2010(4), this data is calculated by author based on quarter-on-quarter GDP growth rate from OECD website. Data on real GDP of Vietnam is obtained from Bloomberg in index form also.
The trade balance is defined as the ratio of exports to imports (X/M). The ratio has been widely used in many empirical investigation of trade balance – exchange rate relationship. It is preferable because it is not sensitive to the unit of measurement and can be interpreted as nominal or real trade balance (Bahmani-Oskooee, 1991). In addition, it neatly solves the problem of using log-form of a trade deficit.
The foreign income variable is based on a weighted average of the indices of real GDP for seventeen trading partners using total trade shares as weights.
51
4.2.2. Econometric characteristics of the data 4.2.2.1. Seasonality:
Climatic as well as institutional factors which occur regularly within the same time of the year (day, week, month, quarter or six-month period) change consumption behaviours, means of production and communication as well as work, production and leisure related activities. When seasonal factors are very present in a series, they need to be disregard to understand the actual trend of the series.
The graphs of four time series, TB, REER, GDP, GDP* in figure 4.1 intuitively indicate that TB, GDP, GDP* have seasonal characteristic. Although we do not test seasonal characteristic of these series, it’s usually that trade balance and output are strongly impacted by seasonal factor. In literature, all previous study on the exchange rate–trade balance relationship, series data of trade balance and output is always seasonally adjusted. Besides, on our estimation, we find that the result using seasonally adjusted series is more significant than the result using series without seasonal adjustment.
Figure 4.1 – Series used in empirical analysis: TB, REER, GDP,GDP*
0.5 0.6 0.7 0.8 0.9 1.0
2000 2002 2004 2006 2008 2010
TB
75 80 85 90 95 100 105 110 115
2000 2002 2004 2006 2008 2010
REER
60 80 100 120 140 160 180 200 220
2000 2002 2004 2006 2008 2010
GDPVN
90 100 110 120 130 140 150 160 170
2000 2002 2004 2006 2008 2010
GDPW
Source: author’s calculation
GDP GDP*
52
Using Census X12 to seasonally adjust the TB, GDP, GDP* series, we have series for estimation as follow.
Figure 4.2 – Seasonally adjusted series: TB, REER, GDP,GDP*
.55 .60 .65 .70 .75 .80 .85 .90 .95
2000 2002 2004 2006 2008 2010
TB_SA
75 80 85 90 95 100 105 110 115
2000 2002 2004 2006 2008 2010
REER
90 100 110 120 130 140 150 160
2000 2002 2004 2006 2008 2010
GDPW_SA
60 80 100 120 140 160 180 200 220
2000 2002 2004 2006 2008 2010
GDPVN_SA
Source: author’s calculation
4.2.2.2. Stationarity
It is necessary to examine the stationary requirement of the four variables as the stationarity characteristic is very important to estimate techniques. In lights of the short time series, two tests are performed to obtain some degree of robustness. In particularly, the study employs the Augmented Dickey-Fuller (ADF) test and the Phillips-Perron (PP) test. The distribution theory supporting the ADF test assumes that the error in the regression is identical and independently distributed. Hence, auto- correlation and heteroskedasticity should not be present in the estimated residuals.
The PP test is a generalization of the ADF procedure that relaxes the restrictions of autocorrelation and heteroskedasticity by undertaking a non-parametric correclation of the t-test. Table 4.1 reports the test results for the four variables, both in levels and in first-differences.
GDP_SA GDP*_SA
53
Table 4.1 – ADF and PP tests results for non-stationarity of variables Variables ADF test
statistic
Critical values at 1%
PP test statistic
Critical value at 1%
TB -3.4324 -3.5924 -3.3860 -3.5924
REER -1.3254 -3.5924 -1.2121 -3.5924
GDP 0.2430 -3.6104 -2.2450 -3.5924
GDP* -0.0666 -2.6240 -0.6596 -3.5924
DTB -8.4671 -3.5966 -10.9696 -3.5966
DREER -8.5292 -3.5966 -8.5129 -3.5966
DGDP -5.3708 -3.6104 -64.6827 -3.5966
DGDP* -2.8089 -3.6055 -6.8161 -3.5966
Note: All variables are in natural logarithms, D is the first difference operator Source: Author’s calculations
Both tests consistently suggest that the trade balance, real exchange rate, domestic output and foreign output are non-stationary in levels (as ADF test statistics and PP test statistics are higher than critical value respectively), but stationary in first- differences (as ADF test statistics and PP test statistics are lower than critical value respectively). This result clears the way for the cointegration analysis below, i.e. for exploring the existence of trade balance relations both in the long (cointegration) and short run (error correction model).
4.3. Model for estimation:
The correlation matrix between variables in table 4.2 below shows that GDP and GDP* are highly correlated to each other (correlation coefficient is 0.98) due to their trend characteristics. Therefore, we have to eliminate one of them out of the model for the purpose of precluding multi-collinearity. We choose to keep GDP because of its importance to trade balance although its correlation coefficient with TB is lower than GDP*’s correlation coefficient with TB.
54
Table 4.2 – Correlation matrix of variables
TB REER GDP GDP*
TB 1.000 0.407 -0.479 -0.525 REER 0.407 1.000 -0.676 -0.659 GDP -0.479 -0.676 1.000 0.980 GDP* -0.525 -0.659 0.980 1.000
Source: author’s calculation
Thus, we come to the final model for estimation as follow:
TBt = + REERt + GDPt + εt (4.2)
4.4. Estimation of long-run effect
In order to explore the existence of a long run relation for trade balance (4.2), the presence of cointegration between the non-stationary I(1) variables must be tested.
While doing that the Johansen cointegration tests (Johansen, 1996) and autoregressive distributed lag (ARDL) approach of Pesaran, Shin, and Smith (2001) will be respectively used. Although Johansen cointegration analysis is widely known, ADRL is often used in study exploring exchange rate-trade balance relationship because it has some advantages to the other. While Johansen cointegration method estimates the long-run relationships within a context of a system of equations, the ADRL method employs only a single reduced form equation (Pesaran and Shin, 1995). The ADRL approach does not involve pre-testing variables, which means that the test on the existence relationship between variables in level is applicable irrespective of whether the underlying regressors are purely I(0), purely I(1) or mixture of both. This feature alone, given the characteristics of the cyclical components of the data, makes the standard of cointegration technique unsuitable and even the existing unit root tests to identify the order of integration are still highly questionable. Furthermore, the ADRL method avoids the larger number of specification to be made in the standard cointegration test. These include decisions regarding the number of endogenous and exogenous variable (if any) to be included, the treatment of deterministic elements, as well as the optimal number of lags to be specified. The empirical results are generally very sensitive to the method and various alternative choices available in the
55
estimation procedure (Pesaran and Smith, 1998). With the ARDL, it is possible that different variables have different optimal lags, which is impossible with the standard cointegration test. Most importantly, the model could be used with limited sample data (30 observations to 80 observations) in which the set of critical values were developed originally by Narayan (2004) by using GAUSS (Jarita Duasa, 2007).
4.4.1. Johansen’s cointegration analysis
Johansen cointegration test technique use two different likelihood ratio test of significance: Trace test and Maximum Eigenvalue test. The trace test tests the null hypothesis of r cointegrating vectors against the alternative hypothesis of n cointegrating vectors. The maximum eigenvalue test, on the other hand, tests the null hypothesis of r cointegrating vectors against the alternative hypothesis of r + 1 cointegrating vectors. Neither of these test statistics follows a chi square distribution in general; asymptotic critical values can be found in Johansen and Juselius (1990) and are also given by most econometric software packages. The critical values used for the maximum eigenvalue and trace test statistics are ased on a pure unit-root assumption.
Table 4.3 – Cointegration Rank Test: Trace and Maximum Eigenvalue Statistics
Hypothesized Eigenvalue Trace Statistic
Max-Eigen Statistic
0.05 Critical Value Trace
Statistic
0.05 Critical Value Max-Eigen
Statistic
Prob.**
Trace Statistic
Prob.**
Max-Eigen Statistic
None * 0.428406 34.52429 22.93240 24.27596 17.79730 0.0018 0.0077
At most 1 0.242131 11.59189 11.36705 12.32090 11.22480 0.0660 0.0472
At most 2 0.005469 4.129906 0.224844 4.129906 4.129906 0.6929 0.6929
Note: There are 2 lag in the VAR model. Trace test indicates 1 cointegrating equation at 0.05 level while Eigenvalue test indicates 2 cointegrating equation at 0.05 level. * denotes rejection of the hypothesis at the 0.05 level. ** James G. Mackinnon, Alfred A. Haug, and Leo Michelis (1999) p- value.
Source: Author’s calculation
The result reported below in table 4.3 does confirm the existence of one cointegrating relation between TB, REER and GDP.
56
The trace test reported in table 4.3 shows that the null hypothesis of no cointegration is rejected, since the trace statistic is larger than the 5% critical value (34.52>24.27). The null stating that there is at most one cointegrating vector cannot be rejected as 11.59<12.32.
A little bit differently, maximum eigenvalue test shows the null hypothesis of no and one (1) cointegration are both rejected, as the max eigenvalue statistics are larger than the 5% critical value (22.93>17.79, 11.36>11.22 respectively) (Table 4.3).
Thus, both tests show the variables do cointegration. We can estimate the corresponding cointegrating equation, and the results read as follows (standard errors in brackets, t-ratios in square brackets):
TB = 0.14REER – 0.18GDP (4.3)
(0.04591) (0.04197) [3.14178] [-4.36599]
In equation (4.3), the estimated cointegration vector is normalized in a way to give a trade balance equation, i.e. coefficient on TB is set to be 1. In order to check if the procedure is justified, it’s necessary to examine whether the trade balance is endogenous while the exchange rate and domestic output are respectively exogenous variables. This turns out to be the case as the cointegrating vector enters the error correction model (ECM) for trade balance below, while it neither enters ECM for exchange rate nor ECM for domestic output. The Granger Causality testing, reported in table 4.4 also suggests that trade balance is endogenous while real exchange rate and domestic output are respectively exogenous variables.
Table 4.4 – Granger Causality Test in Vector AutoRegression (VAR)
Variables TB REER GDP
TB - 3.598760 3.597069
REER 9.516811 - 4.879888
GDP 3.943990 0.162302 -
All 15.03778 3.720458 7.35745
Source: Author’s calculation
57
Lagged exchange rate and GDP significantly affects trade balance (1st column), hence ―granger causing‖ it. On the other hand neither REER nor GDP are Granger caused by respective of other variable (2nd and 3rd column). Thus, the testing result does show that trade balance is endogenous variable while GDP and exchange rate are exogenous variables, confirming that the Johansen’s procedure is justified.
The trade balance equation (4.3) estimated by Johansen’s method indicates impact of exchange rate and trade balance alike to expectation. In the long run, exchange rate has positive relation to trade balance. In particular, 1 percent depreciation of exchange rate will increase trade balance by 0.14 percent. Otherwise, the effect of domestic output on trade balance is negative, 1 percent increase in domestic output will deteriorate trade balance by 0.18 percent.
4.4.2. Autoregressive distributed lag (ARDL) approach
Following the bounds testing approach of Pesaran, Shin, and Smith (2001), we now re-examine the trade balance equation (4.2).
Pesaran, Shin, and Smith (2001) have developed a bounds testing procedure which incorporates the long-run trade balance equation (4.2) into an ECM. This enables simultaneous evaluation of long- and short-run coefficients, which represents one of the main advantages of this approach.
Let Xt = (TBt, REERt, GDPt,) = (TBt, x’t).
Then an ARDL representation of equation (4.2) reads as follow:
Note: ∆ denotes the first difference, t is trend, p is optimal lag length
For estimating the long-run relation, this approach includes two stages.
- The first stage is testing for the existence of a long-run equilibrium relationship (cointegration) between observed variables, in particular the cointegrating vector (λ1, λ2, λ3) with following hypothesis:
(4.4)