Research Article
Cointegration and Causality between Disaggregated Exports and Economic Growth in
ASEAN-4 Nations
Shahrun Nizam Abdul-Aziz*1, Normala Zulkifli2, Norimah Rambeli@Ramli3, Azila Abdul Razak4
1*,2,3,4Faculty of Management and Economics, Sultan Idris Education University, Malaysia
shahrun@fpe.upsi.edu.my1*
Article History: Received: 10 November 2020; Revised: 12 January 2021; Accepted: 27 January 2021; Published online: 05 April 2021
Abstract: This study aimed to examine the relationship between ASEAN-4’s disaggregates exports (i.e., manufactured and primary exports) and economic growth by utilising the time series data over the period from 1982 to 2017. The Johansen-Juselius multivariate procedure was performed to determine the existence of the long-run relationship between variables, while the Granger causality test within VECM was applied to analyse the long-run and short-run causal directions. Prior to that, the unit root test was conducted to examine the series properties of the variables. The empirical results from the Johansen and Juselius Multivariate Cointegration test revealed that there were long-run equilibrium relationships among variables, while the Granger causality test based on VECM found that the ELG hypothesis for manufactured exports was valid for Indonesia in the long-run and short-run, while in the Philippines this hypothesis was only valid for the short-run. On the other hand, in the case of Malaysia and Thailand, both ELG and GLE hypotheses were valid in both long-run and short-run. For each ASEAN-4 nation the results also revealed that physical capital indirectly caused economic growth via the manufactured exports. Nevertheless, in the case of Malaysia and Thailand, it seemed that the reserve effect was likely to happen whereby the economic growth caused the growth of manufactured exports through the increase of the national production. The growth of the manufactured exports due to the reverse effect in turn caused the demand for imports to increase, particularly the imports of intermediate products. As far as the primary exports were concerned, the ELG hypothesis was valid for Thailand in both long-run and short-run, while for Malaysia and Indonesia, this hypothesis was valid respectively in the long-run and short-run. For Thailand, Indonesia and Malaysia, it appeared that in the short run, human capital indirectly stimulated economic growth via primary exports.
Keywords: Export-led Growth, ASEAN-4, Manufactured Exports, Primary Exports, VECM 1. Introduction
The Export-led Growth hypothesis (ELG) suggests that an increase in export would lead to an increase in economic growth. This hypothesis states that the overall growth of a country is not only driven by increasing the amounts of labour and capital within the economy, but also by expanding their exports. As stated by Krueger (1978), Kavoussi (1984) and Ram (1985), export expansion can stimulate the growth of economy through measures such as the specialisation enhancement, full capacity utilisation of the plant size, creating economies of scale and production efficiency through technological development, rising capital formation, employment generation, knowledge spill-overs and increasing the rate of investment. In this aspect, exports are believed to perform as an “engine of growth”. This has attracted the attention of many experts and policymakers to consider export expansion strategy when formulating an appropriate economic strategy for sustaining growth and development particularly in developing nations.
For almost forty years, ELG has dominated development policy. Germany was the first country to adopt the export-led growth idea in 1950s. Japan followed in Germany’s footsteps in the following decade. The import substitution policy adopted in most developing countries began to decline in the 1970s. This was gradually replaced by an export-oriented industrialization (EOI) policy. This change becomes more obvious when both IMF and World Bank played their important role in the promotion and implementation EOI policy particularly in developing countries (Gereffi and Donald, 1990; Haggard, 1990). In the 1970s and 1980s, the Asian Tigers (viz. Taiwan, Singapore, South Korea and Hong Kong) started to adopt export-led growth strategy and it was disseminated further to several Southeast Asian countries (viz. Malaysia, Thailand, the Philippines and Indonesia) in 1980s and 1990s. During this period, the strategy was also adopted by Mexico and Brazil. The implementation of export-led growth strategy has been successful particularly in the Asian Tigers countries and Japan due to the economic integration through trade that exist between those nations during 1970-1999 (Kumar et al, 2020).
Malaysia, Thailand, the Philippines and Indonesia are known as the ASEAN 4 nations. These four countries are some of the developing ASEAN economies which have been growing very rapidly with a widely held view that such growth is export-led. As a matter of fact, a favourable economic performance in these countries begun when these countries shifted from import-substitution industrialization (ISI) strategy to an export-oriented
Shahrun Nizam Abdul-Aziz*, Normala Zulkifli, Norimah Rambeli@Ramli, Azila Abdul Razak
industrialization (EOI) strategy in the late 1980s and early 1990s (Urata, 1994; Lim, 2004). A significant change in the structure of trade in ASEAN-4 has been witnessed whereby the major exports of these economies changed from primary commodities to manufactured products such as machines, transport equipment and miscellaneous manufactured goods (Lim, 2004) and more than 50 percent of ASEAN-4’s exports were manufactured products (Fukuda and Toya, 1995). This phenomenon has generated a lot of attention among experts and policy makers to study the nexus between exports and economic growth in these economies to see/validate whether the strong economic performance in ASEAN-4 nations are export-led or growth-driven. This is vital because the determination of the causal pattern between export and growth has important implications for the policy-maker’s decisions about the appropriate growth and development strategies and policies to adopt.
Countless studies on ELG have been conducted in the context of ASEAN-4. However, it is important to note that most of them gave their focus only on aggregated exports and are very limited studies on ELG which are using disaggregated exports. The major weakness of using aggregated exports when examining the relationship between exports and economic growth is that it limits our understanding of the important differences between dissimilar export category and their influence on economic growth ; this in turn may lead to inauthentic conclusions and implications for policy (Ugochukwu and Chinyere, 2013; Kalaitzi&Cleeve, 2018). Therefore, this study aimed to contribute to the export-growth literature by re-examining the ELG hypothesis for ASEAN-4 using export components of manufactured and primary sectors. Simultaneously, other important variables such as imports, gross capital formation and labour force will be considered in our model. If the variables exhibit cointegration properties, an error correction model would be used in order to draw better statistical inference on causality pattern.
This paper is organised as the following: Section 2 provides a review of the literature on the relationship between exports and economic growth, while the data collection and methodology are described in Section 3. Section 4 focuses on reports and interpretation of the empirical results, while Section 5 provides the conclusion.
2. Literature Review
The neoclassical growth theory states that export is an important activity in generating economic growth and it is also known as the ELG hypothesis. In the first category studies of this hypothesis, the relationship between export and economic growth was examined based on a simple correlation coefficient between these two variables (such as Michaelly (1977), Balassa (1978), Heller and Porter (1978), and Tyler (1981)). In simple terms, these studies found that there is strong support in favour of the ELG hypothesis since export and economic growth are closely linked. This method’s main weakness was that a considerable level of positive correlation between exports and economic growth was used as a foundation to support the ELG hypothesis.
In the second category of studies, the neoclassical growth of production model was adopted by adding exports as an additional regressor. Feder (1983), Kavoussi (1984), and Moschos (1989 stated that the value of coefficient of export growth variable in the growth accounting equation showed highly significant positive correlation. Moreover, according to them, there was substantial improvement in the coefficient of determination in line with the inclusion of export growth in the regression equation. However, some methodological issues still arose whereby a priori assumption had been made which assumed that export causes output growth and did not take into account the direction of causality between the two variables.
In the third category of studies (such as Jung and Marshall (1985), Bahmani-Oskooee et.al. (1991), and Jin and Yu (1995)), the Granger or Sims test of causality was utilised in order to examine the causal direction between export and economic growth. Meanwhile, the recent development in causality test allows researchers to examine both short-and long-run causality between export and economic growth. For instance, Bahmani-Oskooee and Alse (1993) concluded that there was no long-run relationship between exports and economic growth in the Malaysian context, while Dodaro (1993) contended that export growth had a negative effect on the Malaysian economic development. Nevertheless, Doraisami (1996) and Baharumshah&Almasaied (2009) claimed that their results showed a two-way causality between export growth and Malaysian economic growth. In the case of Thailand, Wong (2008) found that there was no cointegration between real GDP and exports. Contrastingly, based on the bounds testing, Jaranyakul (2016) argued that export-led growth hypothesis was valid for Thailand. However, Bahmanee-Oskooee and Alse (1993) found evidence of causation running from economic growth to exports, but no evidence that export caused economic growth in Thailand. Ekanayake (1999) in his study on “Export and Economic Growth in Asian Developing Countries” concluded that there existed bi-directional causal relationship between export and economic growth in Philippines. In contrast, Ahmad and Harnhirun (1996) had reported no relationship between export and growth for Philippines. While Ridzuan et al (2016) contented that the ELG hypothesis had contributed significantly to ASEAN-4 including
Indonesia in the long run. On the other hand, Rahmaddi and Ichihashi (2011) found that exports and economic growth exhibited a bi-directional causal structure, which was ELG in the long-run and growth-led export (GLE) in the short-run concerning the Indonesian case.
In the literature stated above, most researchers accepted that the mixed and inconsistent causality results of the empirical studies on ELG may be due to several reasons such as employed different methods; omission of relevant variables; and arbitrary choice of lags structure. Nonetheless, we added to the literature that the controversial causality results may also be due to the differences in the measure of exports used where this study utilized disaggregated export data (i.e. manufactured export and primary export) in order to test the ELG in the ASEAN-4 nation.
3. Data and Methodology
For this particular study, we utilised the annual data of ASEAN-4 (viz. Malaysia, Thailand, the Philippines and Indonesia) from 1980 to 2017 which was provided by the World Development Indicators (WDI) to analyse the relationship between explanatory variables (i.e., real manufactured exports, real primary exports, real imports, and exchange rate and real foreign income) and economic growth. The dependent of economic growth was measured by real domestic product in current US dollars, while explanatory variables of real manufactured exports and primary in current US dollars were calculated by the author. It was expected that both real manufactured exports and real non manufactured exports to have a positive impact on economic growth because the growth of exports had a stimulating effect on total productivity of the economy as a whole through its positive impact on higher rates of capital (Kavoussi, 1984).
We included additional variable of imports in this study as imports were crucial in testing ELG hypothesis to avoid producing a spurious causality result (Riezmanet al., 1996). It was also pointed out that the finding of no cointegration between exports and output may be due to the omitted variable such as imports. We expected real imports (in current US dollar) to have a positive sign because most of the countries that were involved in this study imported more capital and intermediate goods. Additional variable of exchange (rate official rate) as an endogenous variable was also included as the exchange rate had indirect influence on economic performance via export channel (Al-Yousif, 1999) and we expected this variable to have a positive sign. Foreign income (proxied by real GDP of the US real GDP of the US) as an explanatory variable was also included. The real GDP of the US was chosen for this study as the countries being studied made a huge trade with the US. Foreign income was also expected to have a positive sign because the bigger the foreign income, the bigger demand it can make for imports (Abdul-Aziz etal., 2019).
The VAR Model
The neoclassical growth production function was utilised to generate the following model in this study:
𝐺𝐷𝑃 = 𝑓(𝐺𝐶𝐹, 𝐿𝐹, 𝑋𝑚, 𝑋𝑛𝑚, 𝐼𝑀𝑃) (1)
We considered the unrestricted VAR model below to form the vector autoregressive (VAR) model for this study:
𝑍𝑡 = 𝐴𝑖𝑍𝑡−1+ 𝜀𝑡 𝑘
𝑖=1
(2)
Where𝑍𝑡 contained all the variables of the model in equation (1). The model above can be presented using a
simple six-dimensional VAR model as. 𝐺𝐷𝑃𝑡 𝐺𝐶𝐹𝑡 𝐿𝐹𝑡 𝑋𝑚𝑡 𝑝𝑡 𝐼𝑀𝑃𝑡 = 𝐴0+ 𝐴1 𝐺𝐷𝑃𝑡−1 𝐺𝐶𝐹𝑡−1 𝐿𝐹𝑡−1 𝑋𝑚𝑡−1 𝑋𝑝𝑡−1 𝐼𝑀𝑃𝑡 −1 + 𝐴2 𝐺𝐷𝑃𝑡−2 𝐺𝐶𝐹𝑡−2 𝐿𝐹𝑡−2 𝑋𝑚𝑡−2 𝑋𝑝𝑡−2 𝐼𝑀𝑃𝑡−2 + ⋯ + 𝐴𝑠 𝐺𝐷𝑃𝑡−𝑠 𝐺𝐶𝐹𝑡−𝑠 𝐿𝐹𝑡−𝑠 𝑋𝑚𝑡−𝑠 𝑋𝑝𝑡−𝑠 𝐼𝑀𝑃𝑡−𝑠 (3) where,
Shahrun Nizam Abdul-Aziz*, Normala Zulkifli, Norimah Rambeli@Ramli, Azila Abdul Razak
𝐺𝐷𝑃𝑡 = real gross domestic product
𝐺𝐶𝐹𝑡= gross capital formation
𝐿𝐹𝑡= labour force
𝑋𝑚𝑡= export of manufactured goods
𝑋𝑝𝑡= export of primary goods
𝐼𝑀𝑃𝑡 = real imports
A0= vector of constant terms
Ai= matrices of parameters
Unit root test
Since non-stationary in the variables could lead to spurious regression, a unit root test needed to be conducted in order to test for the presence of a unit root. We carried out both the Augmented Dickey-Fuller (ADF) test and the Phillip-Perron (PP) test to test the stationarity of the series in this study. The ADF test was based on the following three equations:
∆𝑦𝑡 = 𝛾𝑦𝑡 −1+ 𝛽𝑖 𝑝 𝑖=1 ∆𝑦𝑡 −1+ 𝜀𝑡 (4) ∆𝑦𝑡 = 𝛼0+ 𝛾𝑦𝑡−1+ 𝛽𝑖 𝑝 𝑖=1 ∆𝑦𝑡−1+ 𝜀𝑡 (5) ∆𝑦𝑡 = 𝛼0+ 𝛼1𝑡 + 𝛾𝑦𝑡−1+ 𝛽𝑖 𝑝 𝑖=1 ∆𝑦𝑡−1+ 𝜀𝑡 (6)
where 𝑦𝑡= our variable of interest i.e. 𝐺𝐷𝑃𝑡, 𝐺𝐶𝐹𝑡, 𝐿𝐹𝑡, 𝑋𝑚𝑡, 𝑋𝑝𝑡 and 𝐼𝑀𝑃𝑡 ; ∆= differencing operator; 𝑡 =
time trend; 𝜀 = the white noise residual of zero mean and constant variance; 𝛼0 and 𝛼1represented the
deterministic elements. Equations (4), (5) and (6) respectively represented a random walk, a random walk with intercept only and a random walk with intercept and time trend (Gujarati 2003). In each case, the null hypothesis was H0: 𝛾 = 0 (𝑦𝑡is non-stationary / a unit root process), while the alternative hypothesis was Ha: 𝛾 ≠ 0 (𝑦𝑡is
stationary). As such, we could reject the null hypothesis if the t-test statistic from the tests was negatively less than the critical value tabulated.
A non-parametric method of controlling for higher-order serial correlation in a series was suggested by Phillips and Perron (1988). The test regression for the PP test was the AR(1) process:
∆𝑦𝑡= 𝛼 + 𝑦𝑡−1+ 𝜇𝑡 (7)
While the ADF test corrected for higher order serial correlation by adding lagged differenced terms on the right-hand side, the PP test made a correction to the t-statistic of the γ coefficient from the AR(1) regression to account for the serial correlation in 𝜇. The correction was non-parametric since we used an estimate of the spectrum of 𝜇 at frequency zero that was robust to heteroskedasticity and autocorrelation of unknown form. We used Newey-West heteroskedasticity autocorrelation consistent estimate
𝑤2= 𝛾 0+ 2 1 − ( 𝑗 𝑞+ 1)𝛾𝑖, 𝑞 𝑗 =1 𝛾𝑖= 1 𝑇 𝜀𝑡 𝑇 𝑡=𝑗 +1 𝜀𝑡−𝑗 (8)
where𝑞 was the truncation lag. The PP t-statistic was computed as
𝑡𝑝𝑝 = 𝛾01/2𝑡𝑏 𝑤 − (𝑤2− 𝛾 0)𝑇𝑠𝑏 2𝑤𝜎 (9)
where 𝑡𝑏, 𝑠𝑏 were the t-statistic and standard error of 𝛽 and 𝜎 was the standard error of the test regression.
The asymptotic distribution of the PP t-statistic was similar to the ADF t-statistic and we used MacKinnon critical value. We would specify the truncation lag 𝑞 for the Newey-West correction in this test that was the number of periods of serial correlation to include. The dialog initially contained the Newey-West automatic truncation lag selection (the floor function returned the largest integer not exceeding the argument)
𝑞 = 𝑓𝑙𝑜𝑜𝑟 [4(𝑇/100)29] (10)
which was based solely on the number of observations used in the test regression.
The test statistic was the t-statistic for the lagged dependent variable in the test regression for the ADF test. As for the PP test, the test statistic was a modified t-statistic. According to the null hypothesis, a unit root would be rejected against the one-sided alternative if the t-statistic was less than the critical value. This also implied that we failed to reject null hypothesis of a unit root if the t-statistic was greater than the critical value.
Cointegration Test
To carry out the Johansen test, first we considered a VAR of order p:
𝑦𝑡 = 𝐴𝑖 𝑝
𝑖=1 𝑦𝑡 −𝑖+ 𝐵𝜒𝑡+ 𝜀𝑡
(11)
where𝑦𝑡 was a k-vector of non-stationary, I (1) variable, 𝜒𝑡 was a d vector of deterministic variable, and 𝜀𝑡
was a vector of innovations. The VAR would be rewritten as:
Δ𝑦𝑡 = Π𝑦𝑡−1+ Γ𝑡Δ𝑦𝑡−1 𝑝−1
𝑡=1
+ 𝐵𝜒𝑡+ 𝜀𝑡 (12)
where Π = 𝑝𝑡−1𝐴𝑖− 𝐼, Γ𝑖 = − 𝑝𝑡−1𝐴𝑗
Granger’s representation theorem asserts that if the coefficient matrix Π has reduced rank 𝑟 < 𝑘, then there exist 𝑘x 𝑟 matrices 𝛼 and 𝛽 each with rank r such that Π = 𝛼𝛽′ and 𝛽′𝑦𝑡 is stationary. 𝑟is the number of
cointegrating relations (the cointegrating rank) and each column of 𝛽 is the cointegrating vector. As shown below, the elements of 𝛽 are known as the adjustment parameters in the vector error correction model. Johansen’s method is to estimate the Π matrix in an unrestricted form, then test whether we can reject restrictions implied by the reduced rank of Π.
Johansen’s methodology required the estimation of the VAR equation (11) and the residuals were then used to compute two likelihood ratio (LR) test statistics that could be used in the determination of the unique cointegrating vectors of𝑦𝑡. The first test which considered the hypothesis that the rank of Π was less than or
equal to 𝑟 co-integrating vectors was given by the trace test below
𝜆𝑡𝑟𝑎𝑐𝑒 = −𝑇 ln(1 − 𝜆𝑖 𝑘
𝑖=𝑟+1
) (13)
To determine the number of cointegrating relation, 𝑟 subject to the assumption made about the trends in the series, we could proceed sequentially from 𝑟 = 0 to 𝑟 = 𝑘 − 1 until we failed to reject the null hypothesis.
The second test statistic was known as the maximum eigenvalue test which computed the null hypothesis that there were exactly 𝑟 cointegrating vectors in 𝑦𝑡 and was given by
𝜆𝑚𝑎𝑥 = −𝑇𝑙𝑛(1 − 𝜆𝑟) (14)
WhereT was the sample size. Critical values for both tests were tabulated in Osterwald-Lenum (1992). The Short-run Granger Causality Test
Once the variables included in the VAR model were found to be cointegrated, one could use an appropriate error correction models to estimate the long run and short-run causality direction. The set of system equations used in the restricted VAR model (VECM) were shown in the equations (15)-(20) below. In each system equation, the existence of unidirectional long run causal flow could be determined based on the significance and sign of the error correction term, while the existence of the short run unidirectional causality could bedetermined based on the joint significance of the lagged values of the explanatory variables in each system equation. The size of the error correction term indicated the speed of adjustment of any disequilibrium towards a long-run equilibrium state.
Turkish Journal of Computer and Mathematics Education Vol.12 No.3 (2021),
633-645
Research Article Δ𝐺𝐷𝑃𝑡 = 𝛼1+ 𝛽1𝑖 𝑝 𝑖=1 Δ𝐺𝐷𝑃𝑡−𝑖+ 𝛾1𝑖Δ𝐺𝐶𝐹𝑡−𝑖 𝑝 𝑖=1 + 𝛿1𝑖 𝑝 𝑖=1 Δ𝐿𝐹𝑡−𝑖 + 𝜂1𝑖 𝑝 𝑖=1 Δ𝑋𝑚𝑡−𝑖 𝜃1𝑖 𝑝 𝑖=1 𝛥𝑋𝑝𝑡−𝑖+ 𝜉1𝑖 𝑝 𝑖 Δ𝐼𝑀𝑃𝑡−𝑖+ 𝜙1𝐸𝐶𝑇𝑡−1+ 𝜀1𝑡 (15) Δ𝐺𝐶𝐹𝑡 = 𝛼2+ 𝛽2𝑖 𝑝 𝑖=1 Δ𝐺𝐷𝑃𝑡−𝑖+ 𝛾2𝑖Δ𝐺𝐶𝐹𝑡−𝑖 𝑝 𝑖=1 + 𝛿2𝑖 𝑝 𝑖=1 Δ𝐿𝐹𝑡−𝑖+ 𝜂2𝑖 𝑝 𝑖=1 Δ𝑋𝑚𝑡−𝑖 + 𝜃2𝑖 𝑝 𝑖=1 Δ𝑋𝑝𝑡−𝑖+ 𝜉2𝑖 𝑝 𝑖 Δ𝐼𝑀𝑃𝑡−𝑖+ 𝜙2𝐸𝐶𝑇𝑡−1+ 𝜀2𝑡 (16) Δ𝐿𝐹𝑡 = 𝛼3+ 𝛽3𝑖 𝑝 𝑖=1 Δ𝐺𝐷𝑃𝑡−𝑖+ 𝛾3𝑖Δ𝐺𝐶𝐹𝑡−𝑖 𝑝 𝑖=1 + 𝛿3𝑖 𝑝 𝑖=1 Δ𝐿𝐹𝑡−𝑖+ 𝜂3𝑖 𝑝 𝑖=1 Δ𝑋𝑚𝑡−𝑖 + 𝜃3𝑖 𝑝 𝑖=1 Δ𝑋𝑝𝑡−𝑖+ 𝜉3𝑖 𝑝 𝑖 Δ𝐼𝑀𝑃𝑡−𝑖+ 𝜙3𝐸𝐶𝑇𝑡−1+ 𝜀3𝑡 (17) Δ𝑋𝑚𝑡 = 𝛼4+ 𝛽4𝑖 𝑝 𝑖=1 Δ𝐺𝐷𝑃𝑡−𝑖+ 𝛾4𝑖Δ𝐺𝐶𝐹𝑡−𝑖 𝑝 𝑖=1 + 𝛿4𝑖 𝑝 𝑖=1 Δ𝐿𝐹𝑡−𝑖+ 𝜂4𝑖 𝑝 𝑖=1 Δ𝑋𝑚𝑡−𝑖 + 𝜃4𝑖 𝑝 𝑖=1 Δ𝑋𝑝𝑡 −𝑖+ 𝜉4𝑖 𝑝 𝑖 Δ𝐼𝑀𝑃𝑡−𝑖+ 𝜙4𝐸𝐶𝑇𝑡−1+ 𝜀4𝑡 (18) Δ𝑋𝑝𝑡 = 𝛼5+ 𝛽5𝑖 𝑝 𝑖=1 Δ𝐺𝐷𝑃𝑡 −𝑖+ 𝛾5𝑖Δ𝐺𝐶𝐹𝑡−𝑖 𝑝 𝑖=1 + 𝛿5𝑖 𝑝 𝑖=1 Δ𝐿𝐹𝑡−𝑖+ 𝜂5𝑖 𝑝 𝑖=1 Δ𝑋𝑚𝑡−𝑖 + 𝜃5𝑖 𝑝 𝑖=1 Δ𝑋𝑝𝑡−𝑖+ 𝜉5𝑖 𝑝 𝑖 Δ𝐼𝑀𝑃𝑡−𝑖+ 𝜙5𝐸𝐶𝑇𝑡−1+ 𝜀5𝑡 (19) Δ𝐼𝑀𝑃𝑡 = 𝛼6+ 𝛽6𝑖 𝑝 𝑖=1 Δ𝐺𝐷𝑃𝑡−𝑖+ 𝛾6𝑖Δ𝐺𝐶𝐹𝑡−𝑖 𝑝 𝑖=1 + 𝛿6𝑖 𝑝 𝑖=1 Δ𝐿𝐹𝑡−𝑖+ 𝜂6𝑖 𝑝 𝑖=1 Δ𝑋𝑚𝑡−𝑖 + 𝜃6𝑖 𝑝 𝑖=1 Δ𝑋𝑝𝑡 −𝑖+ 𝜉6𝑖 𝑝 𝑖 Δ𝐼𝑀𝑃𝑡−𝑖+ 𝜙6𝐸𝐶𝑇𝑡−1+ 𝜀6𝑡 (20)As seen in the set of equations (15)-(20), 𝐺𝐷𝑃, 𝐺𝐶𝐹𝑚, 𝐿𝐹, 𝑋𝑚, 𝑋𝑝 and 𝐼𝑀𝑃 represented the logged of response and explanatory variables, while 𝛽, 𝛾, 𝛿, 𝜂, 𝜃, 𝜉 and 𝜙 were the regression parameters. ∆was the first difference operator and ECT was the error correction term obtained from the cointegration test. 𝑝was lag length in the autoregressive model and 𝜀 were white noise disturbance.
4. Empirical Results Unit Root Test
In order to assess whether the time series data were stationary or non-stationary, both Augmented Dickey-Fuller (ADF, 1979) and Phillips-Perron (PP, 1988) tests were used to infer the order of integration for the log level and the log of the first difference of each variable. The Akaike Information Criterion (AIC) and/or Schwarz Bayesian Criterion (SBC) determined the lag length for the ADF test. The Newey and West (1979) method was utilised in choosing the truncation lag for the PP test to ensure residuals were white noise. Both tests set the null hypotheses of non-stationarity (existence of unit root) against the alternative of stationarity (no unit root).
Tables 1 and 2 respectively present the results of the ADF and PP tests for Malaysia, Thailand, the Philippines and Indonesia at level and first difference. The ADF and PP results revealed that in each country, the null hypothesis of non-stationary for all the variables failed to be rejected at 5% significance level. On the other
hand, after first differencing the null hypothesis of unit root in each country could be rejected at least at 5% level. As such, we can conclude that all these time series were integrated of order one, I(1).
Table 1.Results of ADF and PP Unit Root Tests (Level) Country Variable ADF PP Constant without Trend Constant with Trend Constant without Trend Constant without Trend Malaysia GDP GCF LF Xm Xnm IMP -1.034 -0.790 0.053 -0.187 -0.770 -2.229 -2.046 -2.186 -1.985 -0.773 -2.130 1.256 -1.052 -0.852 0.053 -2.274 -0.767 -2.611 -2.045 -2.358 -2.146 -0.842 -2.207 -1.643 Thailand GDP GCF LF Xm Xnm IMP -1.277 -1.586 -1.322 -1.023 -0.575 -1.792 -3.556 -1.833 0.078 -0.806 -2.061 -1.543 -1.248 -1.586 -1.564 -1.005 -0.585 -1.780 -1.389 -2.033 0.068 -0.885 -2.412 -1.818 Philippines GDP GCF LF Xm Xnm IMP 0.875 0.518 1.947 1.838 -0.135 -2.229 -2.791 -3.114 -0.589 -0.334 -2.768 -0.101 0.566 0.518 1.145 -1.594 0.212 -2.611 -2.904 -3.189 -2.561 -0.647 -2.769 -0.103 Indonesia GDP GCF LF Xm Xnm IMP 0.024 -0.082 -1.231 2.502 -0.278 -1.905 -2.435 -1.716 -2.348 2.222 -2.525 -2.037 0.043 -0.169 -1.456 2.946 -1.331 -1.941 -2.462 -1.877 -2.489 -3.151 -2.658 -2.281
Table 2.Results of ADF and PP Unit Root Tests (First Difference) Country Variable ADF PP Constant without Trend Constant with Trend Constant without Trend Constant without Trend Malaysia GDP GCF LF Xm Xnm IMP -4.911*** -5.049*** -4.456*** -3.695*** -5.771*** -4.142*** -4.854*** -0.975*** -4.307*** -4.632*** -5.677*** -4.283*** -4.880*** -5.037*** -4.456*** -3.686*** -5.771*** -3.934*** -4.802*** -4.962*** -4.307*** -4.632*** -5.676*** -4.137** Thailand GDP GCF LF Xm Xnm IMP -3.542** -4.294*** -3.172** -3.565** -5.640*** -4.372*** -3.575** -4.228** -5.480*** -5.047*** -5.624*** -4.441*** -3.589** -4.316*** -2.995** -3.785*** -5.641*** -4.277*** -3.621** -4.227** -5.479*** -5.035*** -5.624*** -4.280*** Philippines GDP GCF LF Xm Xnm IMP -4.566*** -4.436*** -4.413*** -4.461*** -5.523*** -4.143*** -4.465**** -4.558*** -4.073** -4.876*** -5.628*** -4.283*** -4.541*** -4.415*** -14.876*** -4.572*** -5.974*** -3.934*** -4.434*** -4.558*** -19.341*** -4.922*** -6.387*** -4.137** Indonesia GDP GCF -5.890*** -4.879*** -5.850*** -4.882*** -5.891*** -4.889*** -5.850*** -4.893***
Note: The null hypothesis (H0) is that each series is I (1). The critical values for rejection of the ADF test are
based on MacKinnon (1996) critical values. Figures in parentheses represent the number of lag structure based on AIC and/or SBC.
Shahrun Nizam Abdul-Aziz*, Normala Zulkifli, Norimah Rambeli@Ramli, Azila Abdul Razak LF Xm Xnm IMP -3.483** -4.062*** -5.033*** -4.568*** -3.756** -5.170*** -4.963*** -4.543*** -3.390** -4.063*** -5.035*** -4.528*** -3.618** -5.155*** -4.967*** -4.508*** Cointegration Test
As all variables in the system were I(1) , we carried out the Johansen test of cointegration to determine the existence of a long-run relationship between the variables in each country under consideration. Table 3 summarised the results for the cointegration test for Malaysia, Thailand, the Philippines and Indonesia. Using the trace and maximum eigenvalue statistics in Table 3, there existed one cointegrating vectors in the model of Indonesia, two cointegrating vectors in the model of Malaysia and the Philippines, and three cointegrating vectors in the model of Thailand. This interpretation showed that all the variables in the Thailand model had a long-run equilibrium relationship with one another and they were adjusting in the short-run via three identified channels. The variables in the system of Malaysia, the Philippines and Indonesia, however, did the same adjustment but only through two channels for both Malaysia and the Philippines, and one channel for Indonesia.
Table 3.Johansen’s Cointegration Test Results Country Cointegrating
Vectors
Trace Max-Eigen
Statistics Critical Value (5%) Statistics Critical Value (5%)
Malaysia r=0 r≤1 r≤2 r≤3 134.735*** 78.437*** 33.345 15.594 95.753 69.818 47.856 29.797 56.297*** 45.092*** 17.750 9.816 40.077 33.876 27.584 21.131 Thailand r=0 r≤1 r≤2 r≤3 146.557*** 93.404*** 57.047*** 26.824 95.753 69.818 47.856 29.797 53.152*** 36.356** 30.223** 17.181 40.077 33.876 27.584 21.131 Philippines r=0 r≤1 r≤2 r≤3 167.004*** 100.619*** 57.978 31.379 95.753 69.818 58.856 29.197 66.384*** 42.640*** 26.598 18.492 40.077 33.876 27.584 21.131 Indonesia r=0 r≤1 r≤2 r≤3 151.766*** 92.213*** 52.940** 24.286 95.753 69.818 47.856 29.797 59.553*** 39.272** 28.653** 14.795 40.077 33.876 27.584 21.131
Vector Error Correction Model
The cointegration test only determined the long-run relationship among variables. It did not mention the direction of the Granger causality. As such, to examine the direction of the Granger causality in the long run and short run we used the vector error correction model (VECM). The long run and short run causality analysis were achieved using a system of equations 15-20 discussed in section 3.4. Each set of equations was used to test one-way causal flow from explanatory to response variables in the five models as showed in Table 4 for Malaysia, Thailand, the Philippines and Indonesia.
Since a cointegrating vector in the six-variable VAR was used in the cointegration tests, it is wise to estimate models with one error correction term (ECT) included to capture long run relationships. The existence of a unidirectional causality from explanatory to response variables was confirmed by the negative sign of the value of the ECT and its statistical significance at the chosen level of significance. Beyond the analysis of the long-run relationships among the six variables in the system, the short-run dynamics were also explored by performing Note: The null hypothesis (H0) is that each series is I (1). The critical values for rejection of the ADF test are
based on MacKinnon (1996) critical values. Figures in parentheses represent the number of lag structure based on AIC and/or SBC.
*, **, *** denote the rejection of the null hypothesis at 10, 5 and 1% respectively.
Note: This test was conducted under the assumption of no deterministic trend in the data and the cointegrating equations have intercepts. The lag length is selected on the basis of AIC and/or SBC. Critical values are taken from Osterwald-Lenum (1992)
multivariate Granger causality tests for the VECM. Chi square statistics and probability (in parentheses) for Granger causality tests from the VECM specification are presented in Table 4 below. The tables also contain the t-statistics for the ECT from each of the five equations.
Table 4.Granger Causality based on Vector Error Correction Model (VECM)
C o un tr y Dep end ent V a ria b les Short-run differences Error Correction Term χ² -test (p-value)
∆GDP ∆GCF ∆LF ∆Xm ∆Xnm ∆IMP t-test Coeff.
MALAYS IA ∆GDP - 3.167 (0.205) 0.975 (0.614) 5.934 (0.05)b 3.167 (0.205) 3.439 (0.179) 2.720 (0.008)a -2.425 ∆GCF 3.180 (0.206) - 2.355 (0.308) 3.713 (0.156) 1.733 (0.421) 5.869 (0.053)c 2.668 (0.007)a 4.965 ∆LF 3.349 (0.187) 6.508 (0.039)b - 3.443 (0.179) 1.057 (0.589) 1.236 (0.539) -1.122 (0.264) 0.228 ∆Xm 5.811 (0.054)c 5.604 (0.061)c 0.184 (0.912) - 1.648 (0.439) 0.942 (0.624) -1.467 (0.145) -1.037 ∆Xnm 3.384 (0.188) 0.519 (0.771) 16.820 (0.000)a 3.619 (0.164) - 1.921 (0.383) 5.482 (0.000)a 0.444 ∆IMP 7.915 (0.019)b 4.524 (0.104) 2.653 (0.265) 5.601 (0.061)c 3.137 (0.208) - 3.988 (0.000)a 4.050 T H A IL A N D ∆GDP - 1.153 (0.562) 2.319 (0.314) 6.512 (0.039)b 6.316 (0.047)b 1.786 (0.409) 2.201 (0.029)b -1.525 ∆GCF 2.514 (0.129) - 0.001 (0.995) 6.259 (0.044)b 2.034 (0.153) 3.178 (0.204) 0.374 (0.708) 0.131 ∆LF 0.242 (0.886) 0.368 (0.832) - 1.707 (0.426) 7.143 (0.028)b 1.263 (0.532) 0.636 (0.525) 0.024 ∆Xm 7.635 (0.022)b 3.632 (0.163) 0.514 (0.773) - 4.716 (0.095) 3.409 (0.182) -0.149 (0.881) -0.052 ∆Xnm 3.229 (0.199) 0.298 (0.861) 2.079 (0.354) 2.164 (0.339) - 0.006 (0.971) -0.975 (0.331) -0.504 ∆IMP 6.579 (0.037)b 0.285 (0.867) 2.115 (0.347) 3.938 (0.139) 5.635 (0.059)c - 0.865 (0.388) 0.438 P H IL IP P IN E S ∆GDP - 1.336 (0.513) 0.227 (0.893) 15.076 (0.000)a 2.856 (0.239) 14.883 (0.000)a 1.715 (0.089)c 0.210 ∆GCF 1.875 (0.392) - 4.417 (0.109) 0.704 (0.703) 0.554 (0.758) 0.181 (0.914) 2.457 (0.015)b 0.719 ∆LF 1.152 (0.562) 0.434 (0.804) - 0.008 (0.995) 0.405 (0.816) 0.312 (0.855) 2.303 (0.023)b 0.041 ∆Xm 1.220 (0.543) 8.111 (0.017)b 1.626 (0.443) - 0.569 (0.752) 1.539 (0.463) 0.681 (0.497) 0.354 ∆Xnm 2.725 (0.256) 1.187 (0.552) 2.044 (0.359) 2.939 (0.274) - 1.432 (0.488) 0.957 (0.340) 0.851 ∆IMP 3.194 (0.202) 1.767 (0.413) 4.231 (0.121) 2.669 (0.263) 2.118 (0.346) - 1.022 (0.309) 0.292 INDON E S IA ∆GDP - 1.999 (0.157) 0.115 (0.734) 6.007 (0.049)b 5.407 (0.020)b 1.022 (0.312) 3.419 (0.000)a -0.746 ∆GCF 0.429 (0.512) - 0.203 (0.653) 4.122 (0.042)b 2.231 (0.328) 0.252 (0.616) 2.468 (0.015)b 0.653 ∆LF 0.174 (0.677) 1.218 (0.269) - 0.335 (0.563) 0.080 (0.777) 2.096 (0.148) -0.472 (0.637) -0.007 ∆Xm 0.031 (0.860) 6.514 (0.038)b 0.237 (0.626) - 0.862 (0.353) 0.120 (0.728) -1.488 (0.138) -0.214 ∆Xnm 0.002 (0.991) 1.624 (0.203) 3.227 (0.072)c 1.310 (0.519) - 0.831 (0.362) 1.036 (0.301) 0.209 ∆IMP 0.243 (0.622) 1.026 (0.311) 0.034 (0.854) 3.200 (0.074)c 2.385 (0.123) - 2.303 (0.023)b 0.560 Note: The uppercases “a”, “b” and “c” denote statistically significant at 1 percent, 5 percent and 10 percent
Shahrun Nizam Abdul-Aziz*, Normala Zulkifli, Norimah Rambeli@Ramli, Azila Abdul Razak
respectively. Figures in the parentheses are the p-value.
In Malaysia’s case, the results in Table 4 revealed that the ECT in the GDP model had the expected negative sign and was statistically significant at 1% significance level. As such, this indicated a unidirectional causal flow running from the explanatory variables to economic growth (GDP). The unidirectional causality flow from manufactured and primary exports to GDP supported the export-led growth hypothesis in both manufactured and primary sectors in the long-run. The coefficient of ECT also implied that the speed of adjustment for economic growth towards equilibrium was very fast (i.e., about 242.5%) from the past year’s deviation in order to attain stability.
The Granger causality results in the short run revealed that the null hypotheses of both non-causality from manufactured exports to GDP as well as from GDP to manufactured exports could not be rejected respectively at 5% and 10% significance levels (c.f. Table 4). These results showed that in the case of Malaysia, there existed bidirectional causality between manufactured exports and economic growth. Additionally, the results in Table 4 also showed that there was a unidirectional causality running from: (1) gross capital formation to manufactured exports which showed that physical capital contributed to the expansion of manufactured exports; (2) labour force to primary exports which indicated that physical capital contribute to the expansion of primary exports; (3) manufactured exports to imports which indicated that the growth in imports had been driven by growing demand from exports of manufactured goods; (4) GDP to imports; and (5) imports to gross capital formation.
In Thailand’s case, the results in Table 4 showed that the ECT for the GDP model was negative and significant at 5% level. Therefore, we can conclude that ELG hypothesis was valid for Thailand in the long run. Additionally, the coefficient of ECT showed that the speed of adjustment for economic growth towards equilibrium was also very rapid (i.e., about 152.5%) from the past year’s deviation so as to attain stability. In the short run, the null hypotheses of both non-causality from manufactured exports to GDP as well as from GDP to manufactured exports cannot be rejected at 5% significance levels. The results informed us that there existed bidirectional causality between manufactured exports and economic growth.
In Malaysia’s case, the results in Table 4 also revealed that there existed unidirectional causality running from: (1) gross capital formation to manufactured exports; (2) labour force to primary exports; (3) manufactured exports to imports; and (4) GDP to imports. Furthermore, Table 4 also revealed that there existed unidirectional causality running from primary exports to GDP which showed that the ELG hypothesis for primary exports was valid for Thailand.
In the case of the Philippines, there was no evidence for the ELG hypothesis in the long run. This was because the ECT had a positive sign even though it was statistically significant at 10% level of significance (c.f. Table 4). Nonetheless, in the short run there was evidence for the ELG hypothesis for manufacturing exports as the null hypothesis of non-granger causality from manufactured exports to GDP was rejected at 1% level of significance. Additionally, based on results in Table 4, there was also evidence of a unidirectional causality running from gross capital formation to manufactured exports,manufactured exports to imports, and imports to GDP. Compared to the cases in Malaysia and Thailand, there was no evidence for primary exports granger cause economic growth for the Philippines.
As seen in Indonesia’s case, the results in Table 4 suggested that there existed a unidirectional causal flow running from the explanatory variables to GDP. This was because the ECT in the GDP model had the expected negative sign and was statistically significant at 1% significance level.
As such, we can state that in the long run the ELG hypothesis for both manufactured and primary exports were valid for Indonesia. The coefficient of ECT also informed us that the speed of adjustment for economic growth towards equilibrium is about 74.6% from the past year’s deviation so as to attain stability. As indicated in Table 4, in the short run the null hypotheses of non-causality from both manufactured exports and primary to GDP failed be rejected at 5% significance levels. These results revealed that there existed unidirectional causality running from both manufactured exports and primary exports to GDP which showed that the ELG hypothesis for manufactured and primary exports were valid for Indonesia in the short run. Furthermore, there was also evidence for a bidirectional causality from gross capital formation to manufactured exports and unidirectional causality from labour force to primary exports and from manufactured exports to imports. Figure 1 below provides a summary of the short-run causal relationship among variables for the countries under study.
Research Article Malaysia 𝑮𝑫𝑷𝑿𝒎 𝑮𝑪𝑭𝑿𝒏𝒎 𝑰𝑴𝑷𝑳𝑭 Thailand 𝑮𝑫𝑷𝑿𝒎 𝑮𝑪𝑭𝑿𝒏𝒎 𝑰𝑴𝑷𝑳𝑭 Philippines 𝑮𝑫𝑷𝑿𝒎 𝑮𝑪𝑭 𝑰𝑴𝑷 Indonesia 𝑮𝑫𝑷𝑿𝒎 𝑮𝑪𝑭𝑿𝒏𝒎 𝑰𝑴𝑷𝑳𝑭
Figure 1.Short-run Lead-Lag Linkages Summarized from VECM
The robustness of the results of the estimated models was confirmed as the models obeyed the classical linear regression such as the disturbance was normally distributed and showed homoscedasticity, there was no serial correlation and the models were correctly specified.
5. Conclusion
This study sought to examine the relationship between disaggregate exports (i.e., manufactured and primary exports) and economic growth in ASEAN-4 utilising the time series data stemming from 1982 to 2017. The approaches that we used in this study were the Johansen-Juselius multivariate procedure to analyse the long-run relationship and Granger causality test within VECM to establish the causal directions. The empirical results from the Johansen and Juselius Multivariate Cointegration test showed that there existed long-run equilibrium relationships among variables under study in our models. In the short-run causality analysis based on disaggregated exports, this study confirmed that there existed bidirectional causality between manufactured exports and economic growth in Malaysia and Thailand. On the other hand, there was a unidirectional causality running from manufactured exports to economic growth which showed that manufactured exports caused economic growth in the Philippines and Indonesia.
Furthermore, the empirical results also revealed that there was a unidirectional causality running from primary exports to economic growth which indicating primary exports causes economic growth in Thailand and Indonesia in the short run. Contrastingly, in the long run over the period 1982-2017 there was also evidence that both manufactured and primary exports caused economic growth in Malaysia, Thailand and Indonesia.
Based on the data above, one can conclude that as far as the manufactured exports are concerned, the ELG hypothesis was valid for Indonesia in the long-run and short-run, while in the Philippines this hypothesis was only valid for the short-run. Meanwhile, in the case of Malaysia and Thailand both ELG and GLE hypotheses were valid in both long-run and short-run. For ASEAN-4 nations it appeared that the role of physical capital was crucial in stimulating the manufactured exports and then the expansion of manufactured export caused economic growth. However, in the case of Malaysia and Thailand, it seemed that the reserve effect happened whereby the growth of economic in turn seems to cause the manufactured exports to grow through increasing the national production.
The expansion of the manufactured exports in turn caused the demand for imports to increase (particularly imports of intermediate products). Accordingly, the government of ASEAN-4 nations should continue the successful export promotion policy and special focus should be placed on the manufactured exports in order to accelerate economic growth. Concurrently, emphasis should be placed on the investment on physical capital as physical capital seemed indirectly to influence economic growth through the expansion of manufactured exports.
Shahrun Nizam Abdul-Aziz*, Normala Zulkifli, Norimah Rambeli@Ramli, Azila Abdul Razak
As far as primary exports are concerned, the ELG hypothesis was valid for Thailand in both long-run and short-run, while for Malaysia and Indonesia, this hypothesis was valid respectively in the long-run and short-run. In the case of Philippines, however, the ELG hypothesis for primary exports was not valid in both the short-run and long-run. For Thailand, Indonesia and Malaysia it appeared that human capital played a vital role in stimulating primary exports. As human capital is important for the primary exports, the government of such countries should focus on human capital accumulation as well as human capital development to accelerate economic growth as this factor indirectly caused economic growth through the expansion of primary exports. References
1. Abdul-Aziz, S. N., Zulkifli, N., Ramli, N., Abdul Karim, N., A., Zakariya, Z., & Abdul Jalil, N. (2019). The Determinations of East Asia’s Automobile Trade Using a Gravity Model. Research in World Economy, 10(5), 113-128.
2. Ahmad, J., &Harnhirun, S. (1996). Cointegration and Causality between Exports and Economic Growth: Evidence from the ASEAN Countries. The Canadian Journal of Economics, 29(2), S413-S416. 3. Al-Yousif, Y. K. (1999). On the Role Exports in the Economic Growth of Malaysia: A Multivariate
Analysis. International Economic Journal, 13(3), 67-75.
4. Baharumshah, A. Z., &Almasaied, S. W. (2009). Foreign Direct Investment and Economic Growth in Malaysia: Interactions with Human Capital and Financial Deepening. Emerging Markets Finance and Trade 45(1), 90-102.
5. Bahmani-Oskooee, M., &Alse, J. (1993). Export Growth and Economic Growth: An Application of Cointegration and Error-Correction Modeling. The Journal of Developing Areas, 27(4), 535-542. 6. Bahmani-Oskooee, M., Mohtad, H., &Shabsigh, G. (1991). Exports, Growth and Causality in LDCs: A
Re-Examination. Journal of Development Economics, 36(2), 405–415.
7. Balassa, B. (1978). Exports and Economic Growth: Further Evidence. Journal of Development Economics, 5(2), 181-189.
8. Dickey, D. A., & Fuller, W. A. (1979). Distribution of the Estimators for Autorregressive Time Series with a Unit Root. Journal of the American Statistical Association, 74(366), 427-431.
9. Dodaro, S. (1993). Exports and Growth: A Reconsideration of Causality. Journal of Developing Areas, 27(2), 227–244.
10. Doraisami, A. (1996). Export Growth and Economic Growth: A Reexamination of Some Time-Series Evidence ofthe Malaysian Experience. The Journal of Developing Areas, 30(2), 223-230.
11. Ekanayake, E. M. (1999). Exports and Economic Growth in Asian Developing Countries: Cointegration and Error-Correction Models. Journal of Economic Development, 24(2), 43-56.
12. Feder, G. (1983). On Exports and Economic Growth. Journal of Development Economics, 12(1-2), 59-73.
13. Fukuda, S.-i. T., H. (1995). Conditional Convergence in East Asian Countries: The Role of Exports in Economic Growth. In T. Ito & A. O. Krueger (Eds.), Growth Theories in Light of the East Asian Experience. Chicago: University of Chicago Press.
14. Gereffi, G., & Donald, L. W. (1990). Manufacturing Miracles: Paths of Industrialisation in Latin America and East Asia. Princeton, NJ: Princeton University Press.
15. Gujarati, D. N. (2003). Basic econometrics (4th ed.). Boston: McGraw-Hill.
16. Haggard, S. (1990). Pathways from the Periphery: The Politics of Growth in the Newly Industrializing Countries. Ithaca: Cornell University Press.
17. Heller, P. S., & Porter, R. C. (1978). Exports and Growth: An Empirical Re-Investigation. Journal of Development Economics, 5(1), 191-193.
18. Jin, J. C. a. Y., E.S.H. (1995). The Causal Relationship between Exports and Income. Journal of Economic Development, 20(1), 131-140.
19. Jiranyakul, K. (2016). The Validity of the Tourism-Led Growth Hypothesis for Thailand. MPRA Paper No. 72806. National Institute of Development Administration.
20. Kalaitzi, A. S., &Cleeve, E. (2018). Export-Led Growth in the UAE: Multivariate Causality between Primary Exports, Manufactured Exports and Economic Growth. Eurasian Bus Rev, 8(1), 341–365. 21. Kavoussi, R. M. (1984). Export Expansion and Economic Growth: Further Empirical Evidence. Journal
of Development Economics, 14(1), 241-250.
22. Krueger, A. O. (1978). Foreign Trade Regimes and Economic Development: Liberalization Attempts and Consequences. Cambridge, MA: Balinger.
23. Kumar, M., Noman, N., &Begam, A. (2020). Export-Led Growth Hypothesis: Empirical Evidence from Selected South Asian Countries. Asian Journal of Economic Modelling, 8(1), 1-15.
24. Lim, C.-Y. (2004). Southeast Asia: The Long Road Ahead (Second Edition). Singapore: World Scientific Publishing Co. Pte. Ltd.
25. Michaely, M. (1977). Exports and Economic Growth: An Empirical Investigation. Journal of Development Economics, 4(1), 49-53.
26. Moschos, D. (1989). Export Expansion, Growth and the Level of Economic Development: An Empirical Analysis. Journal of Development Economics, 30(1), 93-102.
27. Osterwald-Lenum, M. (1992). A Note with Quantiles of the Asymptotic Distribution of the Maximum Likelihood Cointegration Rank Test Statistics. Oxford Bulletin of Economics and Statistics, 54(3), 461– 472.
28. Phillips, P. C. B., &Perron, P. (1988). Testing For a Unit Root in Time Series Regression. Biometrica, 75(2), 335–346.
29. Rahmaddi, R. (2011). Exports and Economic Growth in Indonesia : A Causality Approach based on Multi-Variate Error Correction Model. Journal of International Development and Cooperation, 17(2), 53-73.
30. Ram, R. (1985). Exports and Economic Growth: Some Additional Evidence. Economic Development and Cultural Change, 33(2), 415-425.
31. Ridzuan, A. R., Mohd Noor, A. H., & Ahmed, E. M. (2016). ASEAN4 Prospective of Export-Led Economic Growth. Journal of Business Management and Economics, 7(1), 1-12.
32. Riezman, R. G., Summers, P. M., & Whiteman, C. H. (1996). The Engine of Growth or Its Handmaiden? A Time Series Assessment of Export-Led Growth. Empirical Economics, 21(1), 77–113. 33. Tyler, W. (1981). Growth and Export Expansion in Developing Countries: Some Empirical Evidence.
Journal of Development Economics, 9(1), 121-130.
34. Ugochukwu, U. S., &Chinyere, U. P. (2013). The Impact of Export Trading On Economic Growth in Nigeria. International Journal of Economics, Business and Finance, 1(10), 327 - 341.
35. Urata, S. (1994). Trade Liberalization and Productivity Growth in Asia: Introduction and Major Findings. The Developing Economies, 32(4), 363-372.
