Does Uncertainty Matter for Trade - Economic Growth Nexus in Indonesia?

(1)

3-26-2021

Does Uncertainty Matter for Trade - Economic Growth Nexus in Does Uncertainty Matter for Trade - Economic Growth Nexus in Indonesia?

Indonesia?

Panky Tri Febiyansah

Research Centre for Economics (P2Ekonomi) – The Indonesian Institute of Sciences (LIPI), [email protected]

Follow this and additional works at: https://scholarhub.ui.ac.id/efi Recommended Citation

Recommended Citation

Febiyansah, Panky Tri (2021) "Does Uncertainty Matter for Trade - Economic Growth Nexus in Indonesia?,"

Economics and Finance in Indonesia: Vol. 67: No. 1, Article 10.

DOI: http://dx.doi.org/10.47291/efi.v67i1.933

Available at: https://scholarhub.ui.ac.id/efi/vol67/iss1/10

(2)

Does Uncertainty Matter for Trade - Economic Growth Nexus in Indonesia?

Panky Tri Febiyansahâ,∗, Bintang Dwitya Cahyonoâ, and Rio Novandraâ

aResearch Centre for Economics (P2Ekonomi) – The Indonesian Institute of Sciences (LIPI)

Manuscript Received: 13 January 2021; Revised: 15 March 2021; Accepted: 26 March 2021

Abstract

This paper aims to test the impact of uncertainty on the causal relationship among exports, imports, and economic growth in Indonesia. The relationship is constructed by examining the presence of FDI-adjusted exports and imports (trade) and the output link using conditional variances-covariances derived from the generalized autoregressive conditional heteroskedastic (GARCH) process in a vector error correction model (VEC-GARCH model). Using evidence in Indonesia, the model exposes the uni-directional nexus from trade performance to trade-adjusted output growth in the absence of uncertainty. The volatility effects are evident in the causal relationship between trade and output. The finding shows that the uncertainty effects hamper the trade-economic growth nexus. Incorporated with the long-run causality, trade still causes output even after containing the contributions of volatility. The significant role of imports highlights the higher demand for intermediate capital products and the inclusion of technology in strengthening economic growth.

Keywords:growth; nexus; trade; uncertainty effects; VEC-GARCH model JEL classifications:C32; F14; F21; F41

1. Introduction

The connection between trade openness, shown by less restrictive exports and imports, and the volatility of economic growth is less recognized, while it is commonly considered that economic growth is related to trade openness under appropriate situations (Frankel & Romer 1999). In situations that promote trade openness, external conditions will intensively expose economies, particularly the small open economies, to vulnerability (Helpman

& Krugman 1985 as cited in Mahadevan & Suardi 2008). This produces a higher volatility in the host country that leads to greater instability in economic activities. Thus, it is rational to argue that the presence of uncertainty has the potential to exert influence on the trade-growth nexus.

There is a general perception that exports and im-

∗Corresponding Address: Research Centre for Economics (P2Ekonomi) – The Indonesian Institute of Sciences (LIPI).

Email: [email protected].

ports are significant determinants of the production growth of a country. On one hand, exports will lead to an incentive to producers, resulting in increasing production, better use of resource endowment, access to better technology, and domestic employment (Balassa 1978). It is named the export-led growth hypothesis. On the other hand, imports, particularly intermediate inputs, allow transfer of advanced technology to the host country that leads to an increase in production capacity (Mahadevan

& Suardi 2008). However, while the imported capital goods may bring embodied technical knowledge and increase productivity, imported final goods have adverse effects, such as stimulating a competition and delivering a disincentive to the domestic economy (Kalirajan & Paudel 2015).

To a certain extent, a study shows that foreign direct investment (FDI) has a significant and closely related effect on export performance in developing countries. Generally, FDI should be embedded in exports because it is an important factor of ex-

(3)

ports to strengthen capacity. Furthermore, the FDI- adjusted exports actually represent the ability of the economy to improve productivity. More specifically, Temiz & Gökmen (2011) describe that FDI usually is a subsequent part of global firms with positive contributions to the host country, such as:

increasing productivity and transferring technology, managerial-skilled workers, and greater capacity to produce, leading to strengthening export capacity.

It can be inferred that factoring in FDI in exports is inevitable due to the role of exports in developing economy.

There are numerous studies of trade-economic growth nexus, yet the effects of uncertainty are frequently overlooked in the analyses. Cavallo, De Gregorio & Loayza (2008) show that uncertainty in output growth resulting from trade openness is harmful to economic growth and welfare. Another study also proposes the importance of recognizing the presence of volatility when discussing external economies (Martin & Rogers 2000). Thus, it is necessary to measure and include uncertainty in the analysis of trade-economic growth nexus.

Indonesia as a small open economy is an appropriate example for this empirical study due to the extensive rise in economic activities and trade openness. Considering the effect of global financial crisis (GFC) in 2008, the growth rate has shown an ad- mirable performance in particular, namely in the average range of 5 to 6 percent. However, the ma- jority of exports of Indonesia are primary commodities. As the prices for primary commodities show higher volatility than the manufactured goods in international market, uncertainty plays a crucial role in the export earnings of Indonesia. Furthermore, the demand for imports in the form of final goods in Indonesia has been increasing due to the increasing pressure from its large population. It implies that volatility has the potential to exert significant influence both on trade performance and economic growth.

Based on the aforementioned discussion, the key purpose of this paper is to examine the impact of uncertainty on the causal relationship among ex-

ports, imports, and economic growth in Indonesia in a dynamic framework. This study extends the studies by Cavallo, De Gregorio & Loayza (2008), Martin & Rogers (2000), and Mahadevan & Suardi (2008) in introducing the time-varying component to capture uncertainty in the context of a developing country. The contributions and implications of this study are: (1) to provide a more robust approach to distinguish uncertainty by measuring several pos- sible effects of instability on trade and output in a dynamic context; (2) to assess volatility that will be based on the hysteresis effect, defined as the re- maining effects in uncertainty after considering its cause; and (3) to use a dynamic framework that allows the modelling of time-dependent volatility of trade and income in order to have a more precise measurement of causality.

Empirical findings indicate the existence of long-run stability in the exports-imports-output growth relationship, leading to uni-directional causality in trade- output relationship. Additionally, the trade-output relationship is hampered by volatility. This paper is arranged as follows. Section 2 demonstrates a robust review of practical and theoretical literature.

Section 3 presents analytical procedure and model.

Section 4 explains the data set employed in the estimation. The analysis of the estimation results is described in section 5, while conclusion and policy recommendations are presented in section 6.

2. Literature Review

The theoretical review will deliver theoretical consid- erations and empirical results to construct a framework for trade and output nexus. However, studies using a similar purpose in Indonesia are scarce.

Therefore, this leads to the use of empirical findings in other countries as comparable experiences to the case of Indonesia. There are actually a few of studies that explore the trade-output patterns incorporated with uncertainty.

Export performance is generally believed to be a key factor of the growth of production capacity

(4)

and an increase in employment in the economy.

Balassa (1978), Bhagwati (1978), Edwards (1998), and Sannassee, Seetanah & Jugessur (2014) call this occurrence as the export-led growth (ELG) hypothesis. More specifically, exports play a role in boosting the economy through increasing the num- ber of advanced inputs that in turn fostering more open international trade (Romer 1990; Rivera-Batiz

& Romer 1991). To a certain extent, it works for fundamental studies of how the output growth and exports are strongly linked. Ramos (2001) proposes four reasons to explain the existence of ELG. First, the presence of an export multiplier leads to higher export growth that affects the expansion of output and employment. Second, the benefits obtained from foreign trade allows an increase in imported intermediate goods that can raise the potential output of the economic activities. Third, the rapid increase in competition and the volume of exported products result in accelerated technological improve- ment and economies of scale in production. Con- sidering the previous reasons, the fourth is that the observed robust relationship of output and export growth is inferred as the empirical findings in support of ELG.

Several empirical studies exhibit the presence of ELG hypothesis. Doyle (2001), Parida & Sahoo (2007), and Tang & Ravin (2013) indicate that exports and income growth are significantly con- nected in several developing countries, whether the relationship occurs in short run, long run, both, or undefined. Interestingly, the causal relationship exists in several countries and leads to both ELG and growth driven exports (GDE). In contrast to the ELG hypothesis, Bhagwati (1988) highlights that a rise in output commonly delivers a reliable increase in trade, unless the constantly expanding production generates a resistant-trade bias. Moreover, Krugman (1984) also proposes that GDE can be created by productive expansion due to an increase in the certain level of technology and skilled workers. For instance, Segerstrom, Anant & Dinopoulos (1990) apply the hypothesis of life-cycle of goods that become a theoretical foundation for investigat-

ing a north-south trading process, wherein the production expansion drives competitive research and development in north countries that eventually results in innovation. This implies that export and output relationship varies in terms of the direction and has no identical design, meaning that the nexus is originated from the particular characteristics of a country (Bahmani-Oskooee & Economidou 2009).

To a particular extent, several studies consider the presence of foreign direct investment (FDI) incorporated with exports. For instance, Tarzi (2005) and Kinda (2010) show the effect of FDI inflow on production growth in host countries. The significant interaction occurs because FDI generates direct and indirect spillover effects that can expand production activities and capabilities (Bende- Nabende et al 2003). In turn, the spillover effects result in an increase in export growth, leading to higher economic growth (Chakraborty & Basu 2002). Zhang (2001) and Oladipo (2013) also state that FDI and production growth are significantly correlated in open economic countries. However, there is an empirical study that exhibits varied outcomes of the FDI-output linkage in 47 countries, resulting in a convincing nexus in advanced economies rather than middle-income countries (Qi 2007). In support of this result, Attari, Kama

& Attaria (2011) discover that there is no FDI-GDP relationship in Pakistan, supposing exports are absence. Furthermore, Kohpaiboon (2003), Andraz &

Rodrigues (2010), Guru-Gharana & Adhikari (2011), and Febiyansah (2017) also demonstrate that exports incorporated with FDI drive economic growth intensively. It implies that FDI should be embedded in exports when examining the trade-economic growth nexus in order to explain the role of exports in promoting technology and knowledge dispersion.

Another possibility to boost economic growth as the main determinant of development is to increase imports and expand exports. Awokuse (2007) calls this notion as import-led growth (ILG), proposing that imports can improve production growth. Furthermore, a study by Coe & Helpman (1995) based on the endogenous growth approach

(5)

suggests that imports are one of the factors to achieve long-run output growth by allowing domestic companies to import capital goods and advanced technology. Similar to the finding by Coe &

Helpman (1995), MacDonald (1994), and Lawrence

& Weinstein (1999) discover that imports deliver technology transfer, consecutively enhancing domestic research and development process and urg- ing domestic firms to encourage efficiency and ef- fectiveness. In addition, Lin & Li (2002) explain that supply side is more important to strengthen economic activities by increasing input factors and im- proving an efficient economy that may come from imports. However, instead of corresponding to ILG hypothesis, Thangavelu & Rajaguru (2004) infer that imports can create a valuable ‘virtuous’ nexus between economic growth and trade compared to export performance.

The magnitude and direction of the trade- economic growth relationship varies in each country. Mahadevan & Suardi (2008) propose that the differences in the trade and output variances allows different impacts due to the presence of unpredictable elements. This suggestion is strengthened by the findings by Martin & Rogers (2000) that the increase in the significance of the nexus between uncertainty and growth exists immediately. In re- turn, the uncertainty originating from the volatility of variables can influence the stability of the trade and income nexus significantly either in positive or nega- tive ways. Furthermore, there are a decision making and an exogenous disturbance leading to volatility that needs to be considered in terms of growth (Mahadevan & Suardi 2010). However, it is likely an indication that greater openness delivers higher economic exposure to external disturbances that can leads to more extensive uncertainty in trade pattern and output growth. This implies that the investigation of trade-production growth nexus should not be undertaken separately from the time-varying uncertainty.

Regarding time-varying volatility, the hysteresis hypothesis may explain the role of the time path when uncertainty arises. Lee, Pesaran & Pierse (1992)

examines the effects of hysteresis on trade and output growth in order to investigate the presence of time-path dependence. This study provided evidence that a persistent disturbance from previous periods leads to a large volatility rather than other exogenous shocks. Krugman (1986) suggests an explanation for temporary time dependence in trade- output nexus that time-varying economies of scale can contribute to the presence of hysteresis effects on trade. On the other hand, Göcke (2002) exhibits that a hysteresis process is created by a decline in per unit cost of production because of learning impacts, while Giovannetti & Samiei (1996) present overhead costs as the main driver for hysteresis in trade. These learning impacts include improve- ments in skilled labor and expansions of research and development process. As a result, export capacity is also increased. During the course of bi- causal effects among exports, imports, and GDP, the hysteresis effect in a certain period will further strengthen another hysteresis in the following periods.

Several studies demonstrate the nexus among exports, imports, and economic growth in response to dynamic uncertainty. Mahadevan & Suardi (2008) expose a closed link between trade and output growth in Japan and Asian Tigers by employing a tri-variate ECM-GARCH approach. In uncertain condition, Japan shows the existence of ILG only while Hongkong shows both ILG and ELG. Moreover, the bi-causal connection between trade and economic growth holds in the economy of Korea. Nonethe- less, Taiwan shows a different result, namely the absence of relationship between income growth and both exports and imports. Mahadevan & Suardi (2010) also conducts a case study in Singapore by employing the time series model referred to ECM- ARCH, revealing that income uncertainty hampers both trade and economic growth, although there is bi-directional causality between exports and production growth. In addition, the volatility of neither output nor trade causes a two-way causal nexus between imports and GDP growth, while trade uncertainty has an adverse effect on trade-production

(6)

growth.

In order to measure the trade-output relationship more precisely, a time-varying and dependent uncertainty model is needed for the causality test to isolate and manage the impacts on production and trade growth (Mahadevan & Suardi 2010). How- ever, different from Mahadevan & Suardi, this study follows Grier & Perry (1998) in using the GARCH method that is more likely suitable to predict uncertain and volatile elements of variables. This method is able to explain the real economic meaning of volatility that can no longer be explained by standard deviation by employing the conditional variance-covariance of the volatile shocks of the variables.

3. Method

3.1. Approach

This study applied a procedure containing four stages, namely stationarity, cointegration, post- estimation, and granger-causality tests. The purpose is to perform empirical measurement of the relationship among exports (EXFDI), imports (IMP), and economic growth (GDP) in response to dynamic uncertainty. More specifically, the estimation will present the causality of exports, imports, and economic growth that can be separated into two parts of analysis. First, this model investigated the estimation without the effects of uncertainty. Sec- ond, the causal relationship was generated by including a time-varying uncertainty into the analysis to measure the effect of volatility on each variable.

To address this objective, this study adopted the vector error correction (VEC) model to deal with the causality of the trade-growth nexus in terms of both long-run stability exposed by exogeneity test and short-run relationship. Furthermore, the impacts of uncertainty were tested using the VEC-generalized autoregressive conditional heteroskedastic (VEC- GARCH) model.

Granger offers the ‘Granger test’ to assess the causal relation appropriate to the procedure for an

integrated series. However, time series data some- times reveal an adverse behavior called a random walk that causes miss-specified equations when examining causality process (Granger & Newbold 1974). Furthermore, Granger (1988) explains that supposing the random walk is inevitable and creates cointegrated data, the result of Granger test will be unacceptable and ambiguous. Consequently, Granger test needs to be preceded by an error correction estimation to test causality when the data series are not cointegrated (Granger 2010). This implies that a suitable process in testing causality will be achieved by including the error correction procedure and the measurements of stationarity and cointegration.

The stationarity measurement known as unit root test is a routine process to check stationary time series data from any random walks (Dickey & Fuller 1979). This procedure is appropriate because it can avoid the model from having a spurious regression.

In practice, this study used unit root test applied by Elliott, Rothenberg & Stock (1996), a method that modifies standard dickey-fuller process or aug- mented dickey fuller (ADF) into dickey fuller with generalized least squares (DF-GLS). DF-GLS procedure is preferred because it is more robust and runs adequately in small time series data. The DF- GLS procedure is shown as follows:

∆z^◦_t =α+βz^◦_t−1+

n

X

m=1

θm∆z^◦_t−m+t (1)

where∆ =−(L−1)is constructed by typical lag op- erator.z^◦_t is the locally de-trended series data, while εtindicates a white noise. The distinct procedure in DF-GLS is to estimate GLS prior to examining ADF test. Furthermore, Ng & Perron (2001) describe thatβ = 0is the null hypothesis that performs the series in the unit root process. The expected unit root binds the choice of lag length that delivers selected information criteria. To answer the purposes of this study to a certain extent,Z_tis built from three variables as follows:

Zt= [GDPt,(EXFDI)t,IMPt] (2)

(7)

According to the assumption,Z_thas to be a linear combination in the first order degree one or I(1) to eliminate non-stationary data and ensure the first-order integrated variables hold. Afterwards, the VEC model can be built as follows:

∆Zt= y +αβ⁰Zt−1+

n−1

X

m=1

τm∆Zt−m+t (3)

Where Z_t is a k×1 column vector of variables, αis the vector of correction coefficients,β is the cointegrating relations, andεt is a k×1column vector of disturbances.

Based on the presence of stationary series data, cointegrating relationship is significantly important.

Moreover, it is necessary to first examine the long run relations of the estimated variables in the main model. In response to the estimated cointegration,

the procedure of Johansen (1988) is believed to assist VECM by forming independent identically distributed (iid) Gaussian errors and employing maximum likelihood.

For instance, supposing r represents the cointegration rank, a convincingαβ⁰requires0<r<kas a constraint to prevent a non-singular effect. It infers that VEC model can be generated to recognize the suitable cointegration level. Technically, prior to finding a cointegration system, optimal restriction is convenient to search for optimum lag-length that can exhibit appropriate cointegrating rank.

In practice, the constrained r is identified as a null hypothesis assessed by likelihood statistics through trace measurement and the maximum eigenvalue (Johansen 1991). To clearly illus- trate the VEC model, the expanded model for GDP-exports-imports is presented as follows:

∆GDP_t

∆EXFDIt

∆IMP_t =

α₁ 0 0 0 α2 0 0 0 α₃

µ_1,t−1 µ_2,t−1 µ_3,t−1

+

Pn

m=1ϑm,GDPL^GDP Pn

m=1ϑm,EXFDIL^GDP Pn

m=1ϑm,IMPL^GDP Pn

m=1ϑm,GDPL^EXFDI Pn

m=1ϑm,EXFDIL^EXFDI Pn

m=1ϑm,IMPL^EXFDI Pn

m=1ϑm,GDPL^IMP Pn

m=1ϑm,EXFDIL^IMP Pn

m=1ϑm,IMPL^IMP

∆GDPt

∆EXFDIt

∆IMPt

+

1t

_2t 3t

(4)

Where∆is a first difference representation,nrep- resents the lagged order andεi,tis a non-serial cor- relation in errors following the Gaussian white noise process within observed periods. Moreover,ϑ_km,jis a parameter that captures the short-run dynamic effects on exports, imports, and output growth. Specif- ically,jdenotes GDP, the parametersϑm,GDPL^GDP, ϑm,EXFDIL^GDP, andϑm,IMPL^GDPrepresent the impact of exports, imports, and output growth at period m on the current real output growth. Meanwhile,α is the coefficient of cointegrated error in lagged one that measures speed of adjustment of the equation towards long run stability in each regression.

By the assumption,µ_i,t−1must be stationary zero

degree order or stationary at level, able to perform the I(1) lagged significance of the errors on the VEC equation from the cointegrated equation as follows:

GDPt

EXFDI_t IMPt

=

δ1

δ₂ δ3

+

0 τ12 τ13

τ₂₁ 0 τ₂₃ τ31 τ32 0

GDPt

EXFDI_t IMPt

+

µ_1t µ2t

µ3t

(5)

The determinantτ_ij is the parameter of long run stability. To a certain extent, the cointegrated equation requires an additional measurement to verify causality by employing Granger test that is pre-

(8)

sented by t-statistics and F-statistics on first difference lagged GDPt, EXFDIt and IMPt. In addition, Granger causality test essentially needs another lag criteria since it is sensitive to the lag prefer- ence. Thus, Schwarz information criterion (SIC) and Akaike information criterion (AIC) can be applied.

Considering the presence of the error-corrected arrangement, the causality assessment can be validated using two distinct approaches. Firstly, short-run causality can be examined by generat- ing all one-period lags in difference variables zero for joint-null hypothesis. For instance, supposing ϑ1,EXFDIL^GDP = ϑ2,EXFDIL^GDP = 0 is rejected, then the test suggests that exports growth Granger causes economic growth. Secondly, the long-run causality in response to the effect of error correction captures the combination of short run and long run causality, resulting in exogeneity test in the VECM.

For example, ϑ_1,EXFDIL^GDP = ϑ_2,EXFDIL^GDP = αGDP = 0fails to be accepted, then output growth is Granger-caused by exports growth in terms of both short and long periods.

3.2. Uncertainty and Instability

The measurement of the export-import-GDP nexus is partial supposing this model is unsuccessful to capture the dynamic effects of volatility of trade and income. Factoring in the impacts of volatility patterns creates the possibility to examine which dynamic uncertainty of trade growth affects GDP growth shortfall. Similarly, the generalized autoregressive conditional heteroskedastic (GARCH) model can capture the uncertainty effect in the GDP- Export-Import nexus represented by the conditional variance-covariance of these variables.

Factoring in the impacts of uncertainty allows the examination of whether the decline in economic growth comes from unpredictable trade patterns. The volatility surrounding production and the exports-imports components is measured by the conditional variance-covariance of GDP-Exports- Imports using the GARCH model. The GARCH model is more preferable because it has the benefit

of allowing a past conditional variance correlates with a recent one as well as capture the persistence in time-varying indicators. According to the VEC model, this study defines the conditional variance- covariance matrix as follows:

Vt= F⁰F + D⁰t−1⁰_t−1D + E⁰Vt−1E (6) where

V_t=

vGDP,t vGDP,EXFDI,t vGDP,IMP,t

vEXFDI,GDP,t vEXFDI,t vEXFDI,IMP,t

v_IMP,GDP,t vIMP,EXFDI,t v_IMP,t (7) and F is 3×3 dimensional matrix consisting of six constant elements for all[F] = fijwhilef21 = f₃₁ = f₃₂ = 0. Furthermore,V_t is identified only supposingFis restricted. Then,DandEare3×3 dimensional coefficient matrices with[D] = Dijand [E] = E_ijXX for alli,j = 1,2,3. Note that_j,tshould bej,t∼iidN(0,Vt)andt= [GDP,tEXFDI,t⁰_IMP,t. Here, the conditional variance-covariance arrangement in (6) also provides a consideration about the influence of the conditional variance-covariance among, exports (EXFDI), imports (IMP), and income (GDP) growth.

In addition, to capture uncertainty in the VEC model, equation (6) should be included in equation (3). This implies that VEC model incorporated with variance- covariance specification will generate more accu- rate result in representing trade patterns and output nexus in response to the effect of dynamic uncertainty (Mahadevan & Suardi 2008). This model is named VEC-GARCH as shown in the following:

∆Zt= y +αβ⁰Z_t−1+

n−1

X

m=1

τm∆Z_t−m+ϕVt+t (8)

Where ϕ is the coefficient of uncertainty and Vt is the conditional variance-covariance matrix. The extended VEC-GARCH model is as follows:

(9)

∆DPt

∆EXFDIt

∆IMP_t =

α1 0 0 0 α2 0 0 0 α₃

µ_1,t−1 µ2,t−1

µ_3,t−1

+

Pn

m=1ϑm,GDPL^GDP Pn

m=1ϑm,EXFDIL^GDP Pn

m=1ϑm,IMPL^GDP Pn

m=1ϑ_m,GDPL^EXFDI Pn

m=1ϑ_m,EXFDIL^EXFDI Pn

m=1ϑ_m,IMPL^EXFDI Pn

m=1ϑm,GDPL^IMP Pn

m=1ϑm,EXFDIL^IMP Pn

m=1ϑm,IMPL^IMP

∆GDPt

∆EXFDI_t

∆IMPt

+

ϕ^GDP_GDP ϕ^GDP_EXFDI ϕ^GDP_IMP ϕÊXFDI_IMP ϕÊXFDI_EXFDI ϕÊXFDI_IMP

ϕÎMP_GDP ϕÎMP_EXFDI ϕÎMP_IMP

VGDP,t

V_EXFDI,t VIMP,t

+1t

_2t 3t

(9)

Equation (9) is actually indifferent with the VEC model. Furthermore, the factorϕexamines the impact of volatility resulting from the variable i to j. The Granger causality test was employed in the similar process as the previously common VEC model.

3.3. Data Description

The data set consists of quarterly observations of gross domestic product, exports incorporated with FDI, and imports in Indonesia for the Q1:2000–

Q4:2019 period obtained from several databases.

Regarding the underlying problems in the trade- output causality that needs to be solved, the output variable used the adjusted output, namely GDP minus exports, to avoid both type I and II errors (Riezman, Whiteman & Summers 1996). This infers that adjusted GDP is required to establish the role of imports in affecting both exports and output that can emphasize the impact of exports on economic growth while eliminating the tendency of spurious regression. As a result, this study employed three variables containing adjusted gross domestic product (GDP), real exports over FDI (EXFDI) and real imports (IMP). All variables are set in 2010 constant price and logarithm form. The purpose of using data from 2010 is to avoid the effect of changes in exchange rate due to Asian financial Crisis in 1997 and global financial crisis in 2008. The data set was collected from the database of the bureau statistics (BPS) and central bank (BI) of Indonesia.

Figure 1 displays an obvious indication that GDP and IMP have upward trends and mutually increase

each other. This path shows the expansion of integrated economic activities in Indonesia. Moreover, the slightly increasing trend of GDP means that the engine of growth is not only coming from exports but also other factors, such as aggregate consumption and government spending. However, EXFDI moves in opposite direction, namely downward with some volatile behavior. The trend of EXFDI indicates two reasons. First, exports perform steadily while FDI is more attractive. Second, assuming exposed to global innovation, exports have a slightly declin- ing trend and FDI is relatively stable. In Indonesia, exports incorporated with FDI are appropriate to explain the ability of technology transfer to penetrate domestic production that in turn can drive economic growth. In contrast, the determination of FDI can be occasionally associated with the imported capital goods for supporting domestic economy as shown by an increasing trend of aggregate imports.

Figure 2 exposes the growth rates of GDP, exports, and imports in year-on-year (YoY) technique to similarly denote the effects of a seasonal process in every year. The quarterly mean of GDP in terms of YoY is relatively 0.37 percent lower compared to exports and higher than imports, namely 3.92 percent and 0.34 percent respectively. Particularly for the growth of GDP, regarding the impact of global financial crisis (GFC) in 2008, the data series can be classified into pre-GFC period (2000–2007), within GFC shocks (2008–2009), and post-GFC disturbance (2010–2013), presented respectively about 0.31 percent, 0.50 percent, and 0.38 percent.

Furthermore, exports show a relatively remarkable

(10)

Figure 1. Adjusted Gross Domestic Product (GDP), Exports (EXFDI), and Imports (IMP) in Indonesia, 2000:Q1 to 2019:Q4 (Natural Logarithm)

Source: BPS-Statistics Indonesia (2019) and Bank Indonesia (2019)

Figure 2. Year-on-Year (yoy) Growth Rates of Adjusted Gross Domestic Product (GDP), Exports (EXFDI), and Imports (IMP) in Indonesia, 2000:Q1 to 2019:Q4 (Percent)

Source: BPS-Statistics Indonesia (2019) and Bank Indonesia (2019)

(11)

growth within GFC period, namely 10.31 percent, although they present adverse average of -2.63 percent in post-GFC period. In other words, export performance independently decreases whether FDI is constant or increased in a post-crisis period. Mean- while, imports perform surprisingly positive in the post-crisis, showing an average of 0.36 percent. As- suming the consumption of imported final goods is given, the increasing imports may reflect the presence of greater FDI that needs large imported intermediate products to operate. This condition can result in a credible expansionary level of the economic performance of Indonesia, though it remains not in an optimal condition to achieve potential growth.

4. Result

This paper conducted a formal assessment to verify thatGDPt,EXFDIt, andIMPtare stationary. Table 1 shows the estimated outcome of stationarity measurement. The results indicate that all logarithm forms of variables have no unit root on the first difference while the level meets non-stationary specification. It is also evident from Figure 1 that shows upward trends forGDP_tandIMP_tand a decreas- ing trend forEXFDIt. Specifically,GDPt,EXFDIt, andIMPtare stationary in the first difference at 1 percent. These results verify that all variables are considered to have order degree one integration I(1) that leads to a cointegrated relationship.

Table 1. Stationary Tests

Variable Level First difference

GDP -1.151 -3.706*

EXFDI -0.380 -4.883*

IMP -1.072 -5.735*

Source: Estimation result

Note: P-value shows the significance level by * = 1% ; ** = 5% ;

*** = 10%

The next phase of the formal process is to confirm whether the variables are cointegrated or non-cointegrated in terms of the model, using Johansen and Julius cointegration measurement (JJ-cointegration test). Nonetheless, prior to con-

ducting JJ-cointegration test, an assessment should satisfy the optimal lag selection from five information criteria (Likelihood Ratio, Final Prediction Error, Akaike Information Criterion, Hannan and Quinn Information Criterion, and Schwarz Bayesian Infor- mation Criterion. The outcome of lag length selection is presented in Table 2 that proposes lag 4 in examining VEC model.

Table 2. The Criteria for the Selected Lag Length

Lag 0 1 2 3 4

LR 417.07 12.711 61.422 106.68*

FPE 0.0006 3.40E-06 3.60E-06 2.10E-06 6.5e-07*

AIC 11.725 -393.142 -400.891 -458.025 -5.74713*

HQIC 12.093 -371.048 -375.152 -421.257 -5.26914*

SBIC 12.645 -407.849 -336.489 -366.023 -4.5511*

Note: LR, FPE, AIC, HQIC and SBIC stand for likelihood ratio, final prediction error, Akaike information criterion, Hanan-Quinn information criterion and Schwartz Bayesian information criterion, respectively that are shown by (*) as lag selected

It seems a typical feature of error correction structure in equation (1) contains a set of variables in both integrated degree one and degree zero. This leads to a deviation of asymptotic distribution of assessment statistic from the deterministic components over cointegration test. In determining this obstacle, Table 3 presents trace statistic built by Johansen (1995). Trace statistic creates a joint test to expose cointegration rank that shows that this model is appropriate for the deterministic components. The principal process is to obtain optimal rank by referring to trace statistic.

Table 3. Cointegration Test - Johansen Approach

Maximum Eigenvalue Trace Statistic 5 per cent

rank critical value

0 - 518.895 29.68

1 0.33755 205.914 15.41

2 0.23187 0.5429** 3.76

3 0.00712

Note: ** indicates 5 per cent of significance level

The outcomes of Johansen-cointegrated test are shown in Table 3. Optimal eigenvalue and trace statistic suggest that rank 2 fails to accept null hypothesis at 5 percent of significance level. It infers

(12)

that the model has two cointegrating vectors. There- fore, Johansen-cointegrated test verifies that there is a long-run link amongGDP_t,EXFDI_t, andIMP_t meaning that the trend of all variables moves simul- taneously into the equilibrium level.

Regarding the cointegration test, Figure 2 specifically supports the estimation finding that the model has a long-run cointegrated nexus with respect to the output at time series data set. In a specific manner, an error correction term (ECT), frequently called the residual in error correction model, represents the cointegrating process that moves forward to the long-run stability over time. Overall, Figure 3 confirms that the model has cointegrated vectors.

Subsequent to confirming that the cointegrated relationship in the model prevails, the equation in response to the existence of a long-run linkage can be measured as follows:

GDP{t = 4.455 + 0.032EXFDIt+ 0.758IMPt

(t−statistic) (5.39) (0.88) (12.36) (10) Equation (10) indicates the crowding-in outcome of exports and imports over GDP of Indonesia in the long-run, although exports are not statistically significant. In particular, since GDP, EXFDI, and IMP are in log-forms, the expected coefficients of independent variables are interpreted into elasticity rates. The statistically significance level at 1 percent shows that the raise of 1 percent in imports will boost GDP by 0.758 percent, considering exports. On the other hand, the insignificant exports may come from the role of exports, namely other variables with a greater contribution such as consumption and government spending, in terms of growth of accounting structure. In other words, this equation proposes that external trade expressed by imports and exports have created valuable effects on the growth of output. The estimated results for short-run and strong exogeneity causality tests are stated in Table 4. In addition, the expected outcomes of the causality incorporated with uncertainty effects are reported in Table 5.

Regarding the causality test without uncertainty, this study first discussed the estimations from the VEC model in equations (3) and (4) that are not incorporated with volatility coefficients. The short run test in Table 4 verifies that there are several causal relationships in the short-run between trade and output variables. The test shows that exports and imports significantly cause income. Surprisingly, imports are statistically significant in influencing exports.

It is in compliance with the previously explained hypothesis that exports incorporated with a larger FDI can increase the demand for imported capital goods in order to strengthen and expand the business domestically. Furthermore, by mutually examining short-run and long-run relationships in a joint assessment through an exogeneity test, the estimated outcomes show that output growth causes an increase in openness through imports at 5 percent significance level although exports are not influenced. This confirms that the main determinant of exported goods in Indonesia is an international demand. Meanwhile, exports are still significant to affect GDP and there is evidence of a bi-causal relationship between imports and output.

There is an interesting result in which the magnitude of connection from exports to output is smaller than that of imports. This may indicate the level of development in Indonesia that is still at the expansion stage of domestic production in terms of the presence of FDI (Kacaribu et al. 2018). Furthermore, a larger population of Indonesia drives demand for imported final goods increasing.

In actual condition, the VEC model may not be sufficient to examine the real causality due to the absence of dynamic measurement. It is called uncertainty that occurs in a range of periods when trade pattern and output fluctuate. The implication of avoiding volatility, illustrated from the model that ignores the impacts of uncertainty, may be misin- forming. According to Indrawati et al. (2020), series of uncertain events can reduce the resilience of Indonesia economy. Uncertainty surrounding GDP, exports, and imports growth has the potential to de- crease the effect of trade pattern on output growth

(13)

Figure 3. The Cointegrating Relationship Shown by Error Correction Term (ECT) Source: Estimation result

Table 4. Results of Granger Causality Test by Excluding Uncertainty Effects

Without Uncertainty

Null Hyphotesis EXFDI –> GDP GDP –> EXFDI IMP –> GDP GDP –> IMP IMP –> EXFDI EXPDI –> IMP

Short-run Test 10.032** 18.473 17.801* 12.287** 14.409* 50.699

[P-value] 0.0399 0.7638 0.0013 0.0153 0.0061 0.2802

Causality

Exogeneity Test 10.339** 52.162 16.119* 10.083** 74.557 59.246

[P-value] 0.0351 0.2658 0.0029 0.0391 0.1137 0.2049

Note: P-value shows the significance level by * = 1% ; ** = 5% ; *** = 10%

This sign ( –> ) represents null hyphotesis which is ’does not granger cause’

Table 5. Results of Granger Causality Test by Including Uncertainty Effects

With Uncertainty

Null Hyphotesis EXFDI –> GDP GDP –> EXFDI IMP –> GDP GDP –> IMP IMP –> EXFDI EXPDI –> IMP

Short-run Test 0.2811 26.980 10.576* 6.718** 4.838*** 28.813

[P-value] 0.8689 0.2595 0.0051 0.0348 0.0890 0.2368

Causality

Exogeneity Test 6.177** 43.096 12.856* 36.899 4.987*** 5.609***

[P-value] 0.0456 0.1159 0.0016 0.1580 0.0826 0.0605

Note: P-value shows the significance level by * = 1% ; ** = 5% ; *** = 10%

This sign ( –> ) represents null hyphotesis which is ’does not granger cause’

and vice versa.

Another finding shows the estimation outcomes for equations (8) and (9) by taking the effects of uncertainty on GDP, exports, and imports into consideration. The Granger causality and exogeneity tests were conducted in the similar procedure as in the

VEC model. This model is called VECM-GARCH model. The estimation results are highlighted in Table 5. The figure shows that uncertainty barely hits exports to output nexus in the short run. Since Indonesia has been categorized as a small open economy, export of Indonesia can certainly be di-

(14)

Table 6. The Test for Residual Conducts

Diagnostic on Autocorrelation Diagnostic on Normality Adjustment Speed

Lagrange-Multiplier (lag-4) Jarque-Bera ECT

- - -0.040

Statistics 8.295 0.690 -3.960

P - values 0.505 0.708 0.000

Note: Null hypothesis for Lagrange-Multiplier: no autocorrelation at lag order Null hypothesis for Normality: residual is normally distributed

Null hypothesis for (- ECT): The model has no long-run equilibrium

Figure 4. Stability Analysis Source: Estimation result

minished by global dynamics. The estimation outcome for exports causing output to be eliminated while bi-causal nexus between imports and GDP remains. In addition, imports still influence exports significantly.

In response to volatility as well as both short- run and long-run relationships, exogeneity test shows that uncertainty hampers trade and economic growth nexus. The effect of GDP on imports disappears. Bi-causal relationship between exports and imports are started to establish, which is not shown in the nexus without uncertainty. This may indicate that in the long-run, the performance of exports will be more attractive in attracting imported intermediate commodities while exported-based in- dustries closely depend on products of imports, although the significance is merely 10 percent of the critical level.

Table 6 also provides a post-estimation test showing that export-import-GDP model has no serial- autocorrelation, while the residuals are normally distributed. Furthermore, the coefficient of ECT around -0.04 on∆GDPtexpresses the speed of adjustment that is relatively slower in response to simultane- ously drive the variables in converging to long-run equilibrium at 1 percent significance level. For instance, supposing a shock creates variation in output level in the recent period, the deviation will be adjusted by 4 percent to drive back into equilibrium level in the next periods.

To a certain extent, Figure 4 shows that all eigenval- ues are well-placed inside the circle, indicating the existence of specified model. In addition to Table 6, these infer that all important assumptions of the export-import-GDP relationship model hold.

5. Conclusion

This paper attempts to examine the export-import- GDP nexus in Indonesia using data from 2000:01 to 2019:04. By employing the recent advanced modelling for time-series framework, this paper examines the causality among exports, imports, and GDP growth in terms of dynamics of uncertainty.

The findings show that the exports-imports-GDP growth nexus is significantly cointegrated, meaning it has a tendency to obtain long-run stability. In turn, it confirms that the trade-led growth hypothesis holds in the long-run.

The findings in Indonesia confirm that trade pattern plays a substantial role to drive higher economic growth and strengthen the products of Indonesia

(15)

to be more competitive. In addition to the long-run causality, the short-run causality of trade indicates Granger-influenced economic growth, although the contribution of imports lessens relatively. This condition may be in accordance with the general purpose of the government that imports should be lower in the long-run due to a sufficient technology spill-over and large production capacity.

Uncertainty has significant effects on the trade- economic growth nexus. Volatility hinders short-run causality of the trade-output relationship. In other words, it weakens the nexus. However, the trade- economic growth nexus incorporated with volatility still exists in exogeneity conditions even though the magnitude becomes smaller. We can say that the uncertainty test delivers different and lower measures compared to conventional methods. This will deliver proper information in a decision-making process of trade and economic policies.

Avoiding volatility of production and trade pattern in constructing the nexus model delivers biased estimated outcomes, leading to severe inferences for trade (exports and imports)-led growth policies that decision makers cannot ignore. The results also show substantial intuitions of trade-policy inference that the production capacity and quality of Indonesia is still far from sufficient to meet domestic consumption and global demand for products of Indonesia.

This indication also denotes that the economy activities of Indonesia still face challenges. Furthermore, the estimation results of existing volatility indicate that the traded products and economic situation of Indonesia are highly vulnerable in response to international disturbances. All these findings indicate the importance of national policies in maintaining the stability of economic growth for desirable exports and imports growth and the need to have closer attention to uncertainty issues.

This study has been partially complete. The trade composition is at the aggregate level. Product- specific trade should be taken into consideration.

Further research incorporating disaggregated exports and imports into categorized products may capture time-varying effects more precisely.

References

[1] Andraz, JM & Rodrigues, PM 2010, ‘What causes economic growth in Portugal: exports or inward FDI?’,Jour- nal of Economic Studies, vol. 37, no. 3, pp. 267-287. doi:

https://doi.org/10.1108/01443581011061276.

[2] Attari, MIJ, Kamal, Y, & Attaria, SN 2011, ‘The causal link between foreign direct investment (FDI) and economic growth in Pakistan economy’,The Journal of Commerce, vol. 3, no. 4, pp. 61-68.

[3] Awokuse, TO 2007, ‘Causality between exports, imports, and economic growth: Evidence from transition economies’, Economics Letters, vol. 94, no. 3, pp. 389-395. doi:

https://doi.org/10.1016/j.econlet.2006.08.025.

[4] Bahmani-Oskooee, M & Economidou, C 2009, ‘Export led growth vs. growth led exports: LDCs experience’,The Jour- nal of Developing Areas, vol. 42, no. 2, pp. 179-209. doi:

https://doi.org/10.1353/jda.0.0030.

[5] Balassa, B 1978,.’ Exports and economic growth: Fur- ther evidence’,Journal of Development Economics, vol.

5, no. 2, pp. 181-189. doi: https://doi.org/10.1016/0304- 3878(78)90006-8.

[6] Bank Indonesia 2019,Indonesian economic and financial statistics, Jakarta.

[7] Bende-Nabende, A, Ford, JL, Santoso, B, & Sen, S 2003,

‘The interaction between FDI, output and the spillover variables: co-integration and VAR analyses for APEC, 1965- 1999’,Applied Economics Letters, vol. 10, no. 3, pp. 165- 172. doi: https://doi.org/10.1080/1350485022000044057.

[8] Bhagwati, JN 1978,Foreign trade regimes and economic development: Anatomy and consequences of exchange control regimes, Ballinger Publishing Company.

[9] Bhagwati, JN 1988,Protectionism(Vol. 1), MIT Press.

[10] BPS-Statistics Indonesia 2019, Statistics of Indonesia, Jakarta.

[11] Cavallo, EA, De Gregorio, J, & Loayza, NV 2008, ‘Out- put volatility and openness to trade: A reassessment [with comments]’,Economia, vol. 9, no. 1, pp. 105-152. doi:

https://www.jstor.org/stable/40607909.

[12] Chakraborty, C & Basu, P 2002, ‘Foreign direct investment and growth in India: A cointegration approach’, Applied Economics, vol. 34, no. 9, pp. 1061-1073. doi:

https://doi.org/10.1080/00036840110074079.

[13] Coe, DT & Helpman, E 1995, ‘International R&D spillovers’, European Economic Review, vol. 39, no. 5, pp. 859-887.

doi: https://doi.org/10.1016/0014-2921(94)00100-E.

[14] Dickey, DA & Fuller, WA 1979, ‘Distribution of the estimators for autoregressive time series with a unit root’, Journal of the American Statistical As- sociation, vol. 74, no. 366a, pp. 427-431. doi:

https://doi.org/10.1080/01621459.1979.10482531.

[15] Doyle, E 2001, ‘Export-output causality and the role of exports in Irish growth: 1950-1997’, International Economic Journal, vol. 15, no. 3, pp. 31-54. doi:

https://doi.org/10.1080/10168730100000042.

[16] Edwards, S 1998, ‘Openness, productivity and growth:

what do we really know?’,The Economic Journal, vol. 108,

(16)

no. 447, pp. 383-398. doi: https://doi.org/10.1111/1468- 0297.00293.

[17] Elliott, G, Rothenberg, TJ, & Stock, JH 1996, ‘Efficient tests for an autoregressive unit root’,Econometrica, vol. 64, no.

4, pp. 813-836. doi: https://doi.org/10.2307/2171846.

[18] Febiyansah, PT 2017, ‘Indonesia’s FDI–exports–GDP growth nexus: trade or investment-driven?’, Buletin Ekonomi Moneter dan Perbankan, vol. 19, no. 4, pp. 469- 488. doi: https://doi.org/10.21098/bemp.v19i4.696.

[19] Frankel, JA & Romer, DH 1999, ‘Does trade cause growth?’, American Economic Review, vol. 89, no. 3, pp. 379-399.

doi: 10.1257/aer.89.3.379.

[20] Giovannetti, G & Samiei, H 1996, ’Hysteresis in Ex- ports’, CEPR Discussion Papers 1352, Centre for Economic Policy Research. viewed 2 June 2019 from

<https://cepr.org/active/publications/discussion_papers/

dp.php?dpno=1352>.

[21] Göcke, M 2002, ‘Various concepts of hysteresis applied in economics’,Journal of Economic Surveys, vol. 16, no. 2, pp. 167-188. doi: https://doi.org/10.1111/1467-6419.00163.

[22] Granger, CWJ 1988, ‘Some recent development in a concept of causality’,Journal of Econometrics, vol. 39, no. 1-2, pp. 199-211. doi: https://doi.org/10.1016/0304- 4076(88)90045-0.

[23] Granger, CWJ 2010, ‘Some thoughts on the development of cointegration’,Journal of Econometrics, vol. 158, no. 1, pp. 3-6. doi: https://doi.org/10.1016/j.jeconom.2010.03.002.

[24] Granger, CWJ & Newbold, P 1974, ‘Spurious regressions in econometrics’,Journal of Econometrics, vol. 2, no. 2, pp.

111-120. doi: https://doi.org/10.1016/0304-4076(74)90034- 7.

[25] Grier, KB & Perry, MJ 1998, ‘On inflation and inflation uncertainty in the G7 countries’,Journal of International Money and Finance, vol. 17, no. 4, pp. 671-689. doi:

https://doi.org/10.1016/S0261-5606(98)00023-0.

[26] Guru-Gharana, KK & Adhikari, DR 2011, ‘Economet- ric investigation of relationships among export, FDI and growth in China: An application of Toda yamamoto-Dolado- Lutkephol Granger causality test’,Journal of International Business Research, vol. 10, no. 2, pp. 31-50.

[27] Indrawati, SM, Diop, N, Ikhsan, M, & Kacaribu, F 2020,

‘Enhancing resilience to turbulent global financial mar- kets: An Indonesian experience’, Economics and Fi- nance in Indonesia, vol. 66, no. 1, pp. 47-63. doi:

http://dx.doi.org/10.47291/efi.v66i1.683.

[28] Johansen, S 1988, ‘Statistical analysis of co-integration vectors’,Journal of Economic Dynamics and Control, vol.

12, no. 2-3, pp. 231-254. doi: https://doi.org/10.1016/0165- 1889(88)90041-3.

[29] Johansen, S 1991, ‘Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models’,Econometrica: Journal of the Econo- metric Society, vol. 59, no. 6, pp. 1551-1580. doi:

https://doi.org/10.2307/2938278.

[30] Johansen, S 1995,Likelihood-based inference in cointegrated vector autoregressive models, Oxford University Press.

[31] Kacaribu, F, Lumbanraja, AU, Revindo, MD, Sabrina, S, Pricilia, Z, & Natanael, Y 2018, ‘Recent developments in Indonesia’s macro-economy and trade, Q1-2018’,Eco- nomics and Finance in Indonesia, vol. 64, no. 1, pp. 1-24.

doi: http://dx.doi.org/10.47291/efi.v64i1.582.

[32] Kalirajan, K & Paudel, R 2015, ‘India’s trade deficit with China: will Free Trade Agreement (FTA) work for India?’, Global Economy Journal, vol. 15, no. 4, pp. 485-505. doi:

https://doi.org/10.1515/gej-2015-0011.

[33] Kinda, T 2010, ‘Investment climate and FDI in developing countries: firm-level evidence’, World Development, vol. 38, no. 4, pp. 498-513. doi:

https://doi.org/10.1016/j.worlddev.2009.12.001.

[34] Kohpaiboon, A 2003, ‘Foreign trade regimes and the FDI–growth nexus: A case study of Thailand’,The Jour- nal of Development Studies, vol. 40, no. 2, pp. 55-69. doi:

https://doi.org/10.1080/00220380412331293767.

[35] Krugman, PR 1984, ‘Import protection as export promo- tion’, In Kierzkowski, H (ed.),Monopolistic competition in international trade, Oxford University Press.

[36] Krugman, P 1986, ‘Is the Japan problem over?’,NBER Working Paper Series, 1962, National Bureau of Economic Research. doi: 10.3386/w1962.

[37] Lawrence, RZ & Weinstein, D 1999, ‘Trade and growth:

import-led or export-led? Evidence from Japan and Korea’, NBER Working Paper Series, 7264, National Bureau of Economic Research. doi: 10.3386/w7264.

[38] Lee, KC, Pesaran, MH, & Pierse, RG 1992, ‘Persistence of shocks and their sources in a multisectoral model of UK output growth’,The Economic Journal, vol. 102, no. 411, pp. 342-356. doi: https://doi.org/10.2307/2234519.

[39] Lin, JY & Li, Y 2002, ‘Export and economic growth in China:

A demand-oriented analysis’,mimeo.

[40] MacDonald, JM 1994, ‘Does import competition force efficient production?’, The Review of Eco- nomics and Statistics, vol. 76, no. 4, pp. 721-727.

doi: https://doi.org/10.2307/2109773.

[41] Mahadevan, R & Suardi, S 2008, ‘A dynamic analysis of the impact of uncertainty on import-and/or export-led growth:

The experience of Japan and the Asian Tigers’,Japan and the World Economy, vol. 20, no. 2, pp. 155-174. doi:

https://doi.org/10.1016/j.japwor.2006.10.001.

[42] Mahadevan, R & Suardi, S 2010, ‘Dynamic effects of trade and output volatility on the trade- growth nexus: evidence from Singapore’, Journal of Economic Studies, vol. 37, no. 3, 314-326. doi:

https://doi.org/10.1108/01443581011061294.

[43] Martin, P & Rogers, CA 2000, ‘Long-term growth and short- term economic instability’,European Economic Review, vol.

44, no. 2, pp. 359-381. doi: https://doi.org/10.1016/S0014- 2921(98)00073-7.

[44] Ng, S & Perron, P 2001, ‘Lag length selection and the construction of unit root tests with good size and power’,Econometrica, vol. 69, no. 6, pp. 1519-1554. doi:

https://doi.org/10.1111/1468-0262.00256.

[45] Oladipo, OS 2013, ‘Does foreign direct investment cause long run economic growth? Evidence from the Latin

(17)

American and the Caribbean countries’,International Eco- nomics and Economic Policy, vol. 10, no. 4, pp. 569-582.

doi: https://doi.org/10.1007/s10368-012-0225-4.

[46] Parida, PC & Sahoo, P 2007, ‘Export-led growth in South Asia: a panel cointegration analysis’,International Economic Journal, vol. 21, no. 2, pp. 155-175. doi:

https://doi.org/10.1080/10168730701345414.

[47] Qi, L 2007, ‘The relationship between growth, total investment and inward FDI: evidence from time series data’,In- ternational Review of Applied Economics, vol. 21, no. 1, pp.

119-133. doi: https://doi.org/10.1080/02692170601035058.

[48] Ramos, FFR 2001, ‘Exports, imports, and economic growth in Portugal: evidence from causality and cointegration analysis’,Economic Modelling, vol. 18, no. 4, pp. 613-623. doi:

https://doi.org/10.1016/S0264-9993(00)00055-9.

[49] Riezman, RG, Whiteman, CH, & Summers, PM 1996,

‘The engine of growth or its handmaiden?’, Empir- ical Economics, vol. 21, no. 1, pp. 77-110. doi:

https://doi.org/10.1007/BF01205495.

[50] Rivera-Batiz, LA & Romer, PM 1991, ‘Economic integration and endogenous growth’, The Quarterly Jour- nal of Economics, vol. 106, no. 2, pp. 531-555. doi:

https://doi.org/10.2307/2937946.

[51] Romer, PM 1990, ‘Endogenous technological change’, Journal of Political Economy, vol. 98, no. 5, Part 2, pp.

S71-S102. doi: https://doi.org/10.1086/261725.

[52] Sannassee, RV, Seetanah, B, & Jugessur, J 2014, ‘Export- led growth hypothesis: a meta-analysis’,The Journal of Developing Areas, vol. 48, no. 1, pp. 361-385.

[53] Segerstrom, PS, Anant, TC, & Dinopoulos, E 1990, ‘A Schumpeterian model of the product life cycle’,The Ameri- can Economic Review, vol. 80, no. 5, pp.1077-1091.

[54] Tang, TC & Ravin, C 2013, ‘Export-led growth in Cambodia:

An empirical study’,Discussion Paper 03/13, Department of Economics Monash University. viewed 5 December 2019 from <https://www.monash.edu/__data/assets/pdf_

file/0005/925718/export-led_growth_in_cambodia_an_

empirical_study.pdf>.

[55] Tarzi, S 2005, Foreign direct investment flows into developing countries: Impact of location and government policy’, Journal of Social, Political, and Economic Studies, vol. 30, no. 4, pp. 497-516.

[56] Temiz, D & Gökmen, A 2011, ‘Foreign direct investment (FDI) and export relation in Turkey: 1991–2010’,Journal of Transnational Management, vol. 16, no. 3, pp. 157-180.

doi: https://doi.org/10.1080/15475778.2011.596779.

[57] Thangavelu, SM & Rajaguru, G 2004, ‘Is there an export or import-led productivity growth in rapidly developing Asian countries? A multivariate VAR analysis’, Ap- plied Economics, vol. 36, no. 10, pp. 1083-1093. doi:

https://doi.org/10.1080/0003684042000246795.

[58] Zhang, KH 2001, ‘Does foreign direct investment promote economic growth? Evidence from East Asia and Latin America’,Contemporary Economic Policy, vol. 19, no. 2, pp. 175-185. doi: https://doi.org/10.1111/j.1465- 7287.2001.tb00059.x.