The effect of FDI on economic growth in vietnam

Nội dung text: The effect of FDI on economic growth in vietnam

1. THE EFFECT OF FDI ON ECONOMIC GROWTH IN VIETNAM MBA Mai Anh Linh Email: mailinh.mai7@gmail.com Dr Dang Hoang Ha Email: hoangha@tnu.edu.vn, tel:0977058626 MBA Duong Thi Sen senhoa162@gmail.com International school, Thai Nguyen university, Vietnam Abstract The relationship between Foreign Direct Investment (FDI) and economic growth has motivated a voluminous empirical literature focusing on both developed and developing countries. Empirical work on the impact of FDI in host countries suggests that FDI is an important source of capital, complements domestic private investment, is usually associated with new job opportunities, in most of the cases is related to the enhancement of technology transfer and overall boosts economic growth in host countries. This paper examines this issue in the case of Vietnam by applying the bounds testing (ARDL) approach to cointegration for the period from 1990 to 2017. ARDL approach is used to along with ECM to find out the long run relationship and short-run dynamics between the selected variables. The empirical results indicate a strong relationship between FDI and economic growth in Vietnam. On the policy front, the government could stimulate foreign direct investment through incentives to investors, creation of a good macroeconomic environment and a careful utilisation of loose monetary policy to grow the economy. Keywords: Economic growth, FDI, ARDL, Erros correction, Vietnam 1. Introduction There have been many literatures on this topic, most of which suggest that economic prosperity of a country is related to significant inflows of foreign direct investment (FDI). Many researchers have conducted studies to investigate the theories of FDI, various economic variables that influence FDI, the effects of economic integration on the movements of FDI, and the benefits and costs of FDI, The majority of these studies shows that there is a positive causal relationship between FDI and economic growth, in either the short run, or long run, or both. This implies that that the positive effects of FDI is not inevitable, but rather depend on host countries’ ability to absorb, such as a free trade policy, supportive policy and high quality work force (Zhao and Du, 2007). 57
2. Figure 1: FDI inflows of Vietnam (USD) Vietnam has been quite successful in attracting FDI inflows since the inception of economic reform in 1986. According to a report of Foreign Investment Agency (Ministry of Planning and Investment), the total amount of new registered and increased capital, purchase shares of foreign investors reached 20.33 billion USD in the first semester of 2018, 5.7% more than the corresponding period of the preceding year. In the first 6 months, there were 87 countries and territories with investment projects in Vietnam. The studies on FDI in Vietnam has been voluminous, however, most previous studies forcus on factors affecting FDI in Vietnam which assumes the accurateness of the FDI-growth led theory. In this study, the author would like to confirm this assumption by studying by applying the ARDL Bound Test Approach to Cointegration method using time series. This will contribute to the current literature on the subject and provide useful policy implications based on the revealed short-run and long-run relationship between FDI and economic growth in Vietnam. 2. Theory and Methodology The study follows the model by Bellouni (2014) and Fosu (2006) and others. The basic model of this study is based on endogenous growth theory where total production is a function of technology, capital, and labor. FDI is included in the model to represent the externalities and spill-over effects. The variables capital (domestic investment) and labor are major components in production function which determine the level of production. The control variable, trade openness will capture the externalities in relation to international trade and reduce the omitted variable bias. 58
3. Table 1: Description of Variables Variable Description Time Source of data GDP Real GDP is used as a measure of both a 1990- national total output of goods and World bank 2017 services and its total income FDI Inflow of Foreign Direct Investment is used as an explanatory variable to 1990- World bank explain the FDI - economic growth 2017 relationship DI Domestic Investment is investments by 1990- World bank public and non-public sectors 2017 TO The variable trade openness is calculated 1990- taking the summation of exports and World bank 2017 imports as a ratio to the GDP LB The labor force is used as a proxy for 1990- human capital. The total volume of the World bank 2017 labor force is indicated by LB To analyses the long run as well as short run relationship between selected variables, the study applies ARDL Bound Test Approach to Cointegration. Pesaran & Pesaran (1997), Pesaran & Shin (1999) and Pesaran et al. (2001) consecutively built this approach and found this method to be more proficient than other techniques. There are several relative advantages to the ARDL that make it more useful than others. Firstly, the ARDL allows for the integration of the variables regardless of their order and whether they are stationary at I(0) or I(1). Secondly, the ARDL determines a dynamic unrestricted error correction model (UECM) through a linear transformation. The UECM integrates the short-run dynamics with the long-run equilibrium without losing any information over time. The unrestricted error correction model (UECM) of ARDL approach is used to examine the long run and short run relationship through the following setting: ∆lnGDP = 훿0 + 훿1 lnGDPt-1+ 훿2 lnFDIt-1+훿3 lnDIt-1+훿4 lnTOt-1+ 훿5lnLBt-1+ 푞1 푞2 푞3 푞4 ∑푖=1 훼푖 Δ lnGDPt-i + ∑푖=1 훽푖Δ lnFDIt-i + ∑푖=1 훾푖Δ lnDIt-i + ∑푖=1 휇푖Δ lnTOt-I + 푞5 ∑푖=1 휏푖Δ lnLBt-i + 휀푡 whereGDP is Real GDP; FDI is Foreign Direct Investment; 59
4. DI is Domestic Investment; TO is Trade openness; LB is The total volume of the labor force; ∆ is the first difference operator. In order to streamline the data, all variables were converted to natural logarithm. The use of natural logarithm mitigates correlations among the variables. It also helps in reducing heteroscedasticity as it compresses the scale in which variables are measured. The first part of the equation (1) with 훿1, 훿2, 훿3, 훿4, 훿5, refer to the long run coefficients and the second part with 훼, 훽, 훾, 휇, 휏, 휔, refer to the short run coefficients. To implement the Autoregressive Distribution Lag (ARDL) Bound Test Approach to Co-integration, two steps are involved. First, for testing whether cointegration exists between share prices and the explanatory variables of the model, we test null hypothesis (H0) against alternative hypothesis (H1). H0: 훿1 = 훿2 = 훿3 = 훿4 = 훿5 = 0 H1: 훿1 ≠ 훿2 ≠ 훿3 ≠ 훿4 ≠ 훿5 ≠ 0 Null hypothesis (H0) shows that there is no cointegration between variables while alternative hypothesis (H1) illustrate that co-integration exists between variables. Null hypothesis (H0) against alternative hypothesis (H1) is tested using ARDL bound test. ARDL bound approach to cointegration is a non-standard distribution without considering whether variables are integrated at I(0), I(1) or mix order of integration but no variable is integrated at I(2) or higher order. Pesaran et al. (2001) gave two set of critical values i.e., lower bound values and upper bound values. The set of lower bound values assumes that all variables are I(0) and other set of upper bound values assumes that all variables are I(1). These sets provide a band which covers all possible categories of the integrated variables into the I(0), I(1), even fractionally integrated or mix order of integrated. ARDL bound test is based on F- test. Wald test for determination of F-statistic value is used in the study. If the computed F-statistic value is greater than the critical value of upper bound, it rejects null hypothesis (H0) in favor of alternative hypothesis (H1), indicating that there is cointegration between the variables. If the computed F-statistic value is less than the critical value of lower bound, it rejects alternative hypothesis (H1) in favor of null 60
5. hypothesis (H0), indicates that there is no cointegration exists between the variables. If the computed F-statistic is fall between the lower bound and upper bound, the result is inconclusive. Second, after establishing the cointegration, an appropriate lag length of the variables is selected through Akaike Information Criteria (AIC), Schwarz Information Criterion (SIC), and Hannan-Quinn Criterion (HQ). After determination of appropriate lag length of the selected variables, the long run ARDL model for the stock price is estimated as follow: lnGDP = 훿0 + 훿1 lnGDPt-1+ 훿2 lnFDIt-1+훿3 lnDIt-1+훿4 lnTOt-1+ 훿5lnLBt-1+ 푡 The estimated residual series of the long-run model is known as error correction term (ECT). Next, the error correction model associated is estimated with one lagged ECT to obtain the short-run dynamic parameters. The error correction model is based on the re-parameterization of the estimated long-run ARDL model. The negative and significant coefficient obtained for one lagged ECT will establish the presence of cointegration and it also represents the adjustment speed of the disequilibria from the previous period’s shock which converge back to the long run equilibrium in the current period. 3. Results 3.1. Unitroot test Before applying for econometric work, it is a pre-requisite of checking stationary of series under consideration. According to Granger & Newbold (1974), incorrect inferences would be generated working with non-stationary variables. Using Augmented Dicky-Fuller (ADF) unit root test, we checked the order of integration of selected variables. Table 2 exhibits the univariate analysis results of ADF unit root test for selected variables. According to the table 2, only the variable log LDI is stationary on a level I(0). After running the same test taking the first difference, all the variables became stationary except for labor. Hence, the unit root test confirmed that the variables are stationary on different levels I(0) and I(1). If the variables are stationary on different levels, the most widely used methods identifying co-integration among variables such as Engle and Granger (1987). However, since LLB is I(2), it is droped from the model so that futher steps can be performed. It is worth noting that from the literature review, especially from the papers mentioned above, Labor was not significantly effect GDP and FDI, therefore, it does not affect the model to exclude labor. 61
6. Table 2: Stationarity results of the variables – ADF test statistics results Variables t-Statistic Prob.* Results lnGDP -3.349900 0.0823 Non-stationary lnDI -4.747790 0.0008 Stationary lnFDI -2.486998 0.3311 Non-starionary lnTO -0.851097 0.7866 Non-starionary lnLB 2.132905 1.0000 Non-starionary D(lnGDP) -5.390898 0.0002 Stationary D(lnFDI) -3.862214 0.0070 Stationary D(lnTO) -4.295069 0.0026 Stationary D(lnLB) -2.045729 0.5503 Non-starionary D(D(lnLB)) -4.788685 0.0040 Stationary 3.2. ARDL estimates Eviews software is used to run ARDL approach to co-integration and the test result of the model is shown in table 3. The appropriate lag is selected automatically based on Schwarz Criterion and the selected ARDL model is (1,0,0,1). The probability of F-statistics is less than 0.05 for the short-run model indicating its significance. Diagnostic tests such as serial correlation, normality test, and heteroscedasticity test were conducted to determine the validity of the data. The statistical value of heteroscedasticity is 0.625 and the probability level is 0.708 which is greater than 0.05. It is an indication of lack of heteroscedasticity of the model. As shown in the appendix, the short-run model gets through all the diagnostic tests. There is no serial correlation or autocorrelation and error term of the model is also normally distributed. The estimated results of long-run coefficients are shown in Table 4 with F statistic much greater than the upper bound hence the null hypothesis of no co- integration is rejected. Meaning that a long-run relationship exists between the selected variables in the model. The long-run coefficient is an indicator of the long run relationship of the variables with the dependent variable. The sign of the coefficient of all the selected variables are positive in the long-run model except for trade openness, but only log FDI is significance. Economic growth is positively correlated with DI, negatively correlated with trade openness in long run, but not significant. This is an indication of low productivity in domestic sector, and high produtivity in FDI investments. 62
7. Table 3: ARDL estimetion results Dependent Variable: LGDP Selected Model: ARDL(1, 0, 0, 1) Variable Coefficient Std. Error t-Statistic Prob.* LGDP(-1) 0.361606 0.116871 3.094046 0.0057 LDI 0.145381 0.082370 1.764964 0.0928 LFDI 0.139250 0.033417 4.167004 0.0005 LTO -0.332330 0.187504 -1.772390 0.0916 LTO(-1) 0.299218 0.165906 1.803543 0.0864 C 8.794464 2.119854 4.148619 0.0005 @TREND 0.042667 0.013847 3.081424 0.0059 R-squared 0.997732 Mean dependent var 24.68712 Adjusted R-squared 0.997052 S.D. dependent var 0.978237 S.E. of regression 0.053116 Akaike info criterion -2.814245 Sum squared resid 0.056427 Schwarz criterion -2.478287 Log likelihood 44.99230 Hannan-Quinn criter. -2.714347 F-statistic 1466.445 Durbin-Watson stat 1.399706 Prob(F-statistic) 0.000000 Table 4: Estimations of long-run coefficients Variable Coefficient Std. Error t-Statistic Prob. LDI 0.227729 0.116498 1.954794 0.0647 LFDI 0.218126 0.050761 4.297135 0.0004 LTO -0.051867 0.373492 -0.138870 0.8909 EC = LGDP - (0.2277*LDI + 0.2181*LFDI -0.0519*LTO ) F-Bounds Test Null Hypothesis: No levels relationship Test Statistic Value Signif. I(0) I(1) Asymptotic: n=1000 F-statistic 8.549449 10% 3.47 4.45 k 3 5% 4.01 5.07 2.5% 4.52 5.62 1% 5.17 6.36 63
8. The ECM of the selected ARDL model is shown in Table 4. The short-run elasticity of the variable is shown by the D(.) sigh. According to the estimation results, only lnGDP (lag1) lnFDI, lnTO (lag 1) effect GDP in short-run. A negative sign of the error term is the indication of convergence towards the equilibrium. The estimated values of R2 and adjusted R2 are 0.72 and 0.63 respectively indicating that 72 percent of the variation of the dependent variable is explained by the independent variables. Diagnostic tests showed the of lack of autocorrelation between the variables. Table 5: Error correction representation for the selected ARDL model Dependent Variable: D(LGDP) Method: Least Squares Variable Coefficient Std. Error t-Statistic Prob. C 0.016048 0.026196 0.612598 0.5474 D(LGDP(-1)) 0.404240 0.135971 2.972990 0.0078 D(LDI) 0.178965 0.120290 1.487786 0.1532 D(LFDI) 0.113751 0.038618 2.945565 0.0083 D(LTO) -0.065284 0.155402 -0.420101 0.6791 D(LTO(-1)) 0.463729 0.134504 3.447687 0.0027 ECM(-1) -0.772478 0.245404 -3.147783 0.0053 R-squared 0.720142 Mean dependent var 0.121058 Adjusted R-squared 0.631766 S.D. dependent var 0.075028 S.E. of regression 0.045529 Akaike info criterion -3.116152 Sum squared resid 0.039384 Schwarz criterion -2.777434 Log likelihood 47.50998 Hannan-Quinn criter. -3.018614 F-statistic 8.148614 Durbin-Watson stat 1.533212 Prob(F-statistic) 0.000188 3.3. Diagnostic tests Figure 1 shows the CUSUM and CUSUMSQ stability tests respectively for the selected ARDL based ECM. This model is stable because none of the lines cross the critical value lines of figure plots generated by the Eviews software. Hence, this model can be applied to explain FDI economic growth relationship in Vietnam. 64
9. 15 1.6 10 1.2 5 0.8 0 0.4 -5 -10 0.0 -15 -0.4 2000 2002 2004 2006 2008 2010 2012 2014 2016 2000 2002 2004 2006 2008 2010 2012 2014 2016 CUSUM 5% Significance CUSUM of Squares 5% Significance 4. Conclusions and Recommendations The core objective of this study is to find out FDI-economic growth relationship of Vietnam for the period of 1990-2017. The selected variables are real FDI inflows, domestic investment, trade openness to identify the impact on economic growth. Vietnam is seeking FDI to boost economic growth at present, hence, it is motivated identifying whether FDI promotes economic growth on prevailing economic setting. ARDL approach is used to along with ECM to find out the long run relationship and short-run dynamics between the selected variables. The ECM generated expected sign at 1 percent significant level. The selected econometric model gets through all the diagnostic tests and confirms the absence of serial correlation, heteroscedasticity, and nonnormality. CUSUM and CUSUMSQ tests confirm the stability of the model validating the applicability in policy making. The empirical results the study are associated with previous research which found a strong relationship between FDI and economic growth in Vietnam. Among the other selected variables, FDI is found as the main driver of economic growth in Vietnam. DI is weakly and positively correlated with economic growth in the long run and short run. This is an indication of low productivity in domestic sector, and high productivity in FDI investments. Inshort run, GDP and trade openness in the previous period also act as positive indicators for economic growth. 5. References 1. Agosin, M., & Mayer, R. (2000). Foreign direct investment: Does it crowd in domestic investment? United Nations Conference on Trade and Development Working Paper No. 146, Geneva, Switzerland. 2. Ball, D. A., & McCulloch, W. H. (1999). Glossary. International Business: The challenge of global competition (7th ed.). The McGraw-Hill Companies. 3. Belloumi, M. (2014). The relationship between trade, FDI and economic growth in Tunisia: An application of the autoregressive distributed lag model. Economic Systems, 38(2), 269-287. doi:10.1016/j.ecosys.2013.09.002 65
10. 4. Buckley, P. J., Clegg, J., & Wang, C. (2002). The impact of inward FDI on the performance of Chinese manufacturing firms. Journal of International Business Studies, 33(4), 637–655. 5. Creswell, J. W. (2009). The selection of a research design. Research design: Qualitative, Quantitative, and Mixed methods approaches (3rd ed.). Sage Publications, Inc.: Thousands Oaks, CA 6. Fosu, O. E., & Magnus, F. J. (2006). Bounds Testing Approach to Cointegration: An Examination of Foreign Direct Investment Trade and Growth Relationships. American Journal of Applied Sciences, 3(11), 2079-2085. doi:10.3844/ajassp.2006.2079.2085Tran, N. N. A. T., & Le, H. P (2014). Impact of Public Investment on Economic Growth in Vietnam: An Experimental look of ARDL model. Journal of Investments and Integration, 19(29),10-19. 7. Zhao, C., Du, J., 2007. Causality between FDI and economic growth in China. Chinese Econ. 40 (6), 68–82. 66