Chinas Fdi And Malaysias Economic Growth

Published Date: 02 Nov 2017

3.1 Introduction of methodology

In this chapter, we introduced the theoretical framework as well as econometric technique that help us to estimate the relationship between China’s (outward) FDI and Malaysia’s economic growth. With respect to theoretical framework, we introduced two models in this research. First of all, the relationship between China’s FDI, Malaysia’s trade openness, financial development (all of the above are independent variable) and Malaysia’s economic growth (the dependent variable). Secondly, we also hypothesize that Malaysia’s market size, exchange rate and human capital development will affect China’s outward FDI. In this case, the variables that we are going to use contain data from 1987 to 2009.

On the other hand, in terms of econometric techniques, we incorporate the concept of cointegration which was initiated by Granger (1981) and Granger and Weiss (1983) before it was extended and modified by Engle and Granger in 1987. The notion of cointegration explains the existence of a stationary or long-run equilibrium relationship among two or more variables (time series), though they are individually non-stationary (Narayan & Narayan, 2004). The major benefit of cointegration method is that it facilitates us in integrating the short-run and long-run relationships between two or among more than two variables in a framework that is unified. In addition, the spurious/nonsense regression problem can be partially or totally eliminated by the existence of cointegration among the variables. Under the notion of cointegration, we employ the ARDL approach which is also known as the bound testing procedure.

Moreover, in order to test the short term causal relationship between the variables, we also apply Granger causality tests. Before we proceed to cointegration test, we employ unit root tests such as Augmented Dickey Fuller (ADF) test, and Kwiatkowski-Phillips-Schmidt-Shin (KPSS) to examine the stationarity of the data and confirm that the various series suit the requirement of same order of integration with I (d) and linear combination of all variables must be I (d-b).

3.2 Econometric Model

Based on previous studies, we reinforced the model of Malaysia GDP and China’s foreign direct investment (FDI) respectively as follow:

MGDP =F (CFDIM, MOPEN, FD)

CFDIM = F(MMS, EXR, HCD)

3.2.1 Source of Data and Definitions

Variable

Definition

Source of data

MGDP

Gross Domestic Product (GDP) in Malaysia

IFS (International Financial statistic)

CFDIM

China’s foreign direct investment inflows of Malaysia (RM Billion)

MIDA (Malaysian Industrial Development Authority)

MOPEN

Export plus imports divided by nominal GDP (Malaysia’s openness)

IFS (International Financial statistic)

FD

M3 divided by Gross Domestic Product (GDP)

BNM (Bank Negara Malaysia)

MMS

Nominal GDP in Malaysia

(Malaysia’s market size)

IFS (International Financial statistic)

EXR

Yuan Renminbi per Ringgit Malaysia

(RMB/RM)

IFS (International Financial statistic)

HCD

Employment rate in Malaysia (Human capital development)

World Bank

3.3 Econometric Method

In this study, we employ the time series method to estimate our research model, whereby the method is suitable for researches that focus only on one country with many time periods. Furthermore, the unique of time series techniques is its ability to decompose a trend, a seasonal, a cyclical, and an irregular component. The most important aim of time series method is to have forecast based on economic data. This study aims to test on how China’s foreign direct investment (FDI) affects Malaysia’s economic performance, and determine whether what are the factors that can attract China’s FDI into Malaysia.

There are two time series approaches that we employ in the next section:

Bound Test (unrestricted error correction model)

Granger Causality Tests

3.3.1 ARDL approach

To study the long-run relationship between China’s foreign direct investment and Malaysia’s economic growth, we employ the ARDL approach advocated by Pesaran and Shin (1995) (see also Pesaran & Pesaran 1997; Pesaran, Shin, & Smith 2001). The ARDL approach is also known as the bound testing method. The concept of cointegration was first established by Granger (1981) and Granger and Weiss (1983). Eventually, it was further extended and formalized by Engle and Granger (1987). Cointegration illustrates the existence of an equilibrium or stationary relationship among two or more time series, in which each of them is individually nonstationary. Its pro is that it enables us to integrate the long-run and short-run relationships between variables in a standardized framework. After the seminal effort of Engle and Granger (1987), studies on cointegration techniques have been advanced with attention is paid on determining the number of linearly independent cointegration vectors, or the cointegrating rank, in a typical vector autoregressive process.

The bound testing approach has the underlying computed F-statistic (Wald F-statistic), which is used to examine the significance of lagged levels of the variables under the study in a conditional unrestricted equilibrium correction model (UECM). This approach has a few pros compared to others like Johansen and Juselius (JJ) tests, Engle-Granger two-step method. One of the benefits is that regardless of the stationarity of the independent variables (whether they are in I(0), or I(1), or mutually cointegrated), the ARDL method is applicable. Therefore, due to its independency on pretesting the variables’ order of integration, bound testing approach eliminates the risks related to pre-testing the order of integration. Furthermore, bound testing procedure can avoid the finite sample size problem which is suffered by JJ tests and Engle-Granger two-step method. In other words, it is reliable to be employed in research that involves small sample size as compared to JJ tests and Engle-Granger two-step procedure.

In order to employ bound testing procedure, we demonstrated the Vector Auto-Regression (VAR) of order p(VAR(p)) for China’s FDI-led growth function in Malaysia.

(1)

where zt is defined as the vector of both yt and xt, where yt is the endogenous variable (measured by Malaysia’s economic growth and China’s FDI in Malaysia) and the xt (CDIM, MOPENNESS, FD, MMS, EXR, HCD) is the vector matrix of a set of exogenous variables. ‘t’ is a time or trend variable. According to Pesaran et al. (2001), yt must be I(1) variable, but the independent variables, xt can be either I(0) or I(1).

VECM (Vector Error Correction Model) can be further developed as below:

(2)

Where Δ=1-L. The long run multiplier matrix as developed by Choong et al. (2005) are now separated as:

(3)

The crossway components of the matrix are unrestricted. Thus, the chosen series can be either I(0) or I(1). If λyy 0, then y is I (0). Contradictory, if λyy = 0, then y is equal to I (1).

The ARDL method involves two stages (Narayan & Narayan, 2004). Establishing the existence of a long run relationship is the first stage. After the long run relationship has been established, a two-step procedure is employed to estimate the long-run relationship. To investigate whether a long-run relationship is present in equation 1 or not, we have to estimate the following unrestricted error correction (UEC) models:

(4)

(5)

The existence of long-run relationship is tested by using F-test, whereby the F-test will indicate whether which variable should be normalized when there is a long run relationship in the model (Narayan & Narayan, 2004). Besides, the null and alternative hypotheses for the respective models are constructed as the following:

H0:ψ1x= ψ 2x=0 (no long run levels relationship)

H1: ψ 1x ψ 2x0 (long run levels relationship exists)

(From equation 4)

H0:δ1Y=δ2Y=0 (no long run levels relationship)

H1:δ1Yδ2Y0 (long run levels relationship exists)

(From equation 5)

This can also be indicated as FX (X|Y) for equation 4 and FY (Y|X) for equation 5.

The F-test follows a non-standard distribution which depends on (i) whether the UEC model is with drift and/ or a trend or not (ii) the number of exogenous variables, and (iii) whether the variables included in the UEC model are of one order of integration I(1) or zero order of integration I(0) (Narayan & Narayan, 2004). In Pesaran and Pesaran (1997) as well as Pesaran et al. (2001), there were two sets of critical values being reported. Unfortunately, they are not suitable for us due to our finite sample size (23 observations only). Hence, we have to calculate critical values based on our sample size, generated by Narayan (2005). We carried out a further two-step procedure in the second stage, to estimate the model once a cointegration has been established. First of all, we approached adequate lag information criteria such as Schwartz Bayesian Criteria (SBC) to select the (optimal) order of the lags in the model (ARDL). Then, we employed OLS method to estimate the chosen model.

In order to examine the impact of China’s foreign direct investment (FDI) on Malaysia’s economic growth and the determinants of China’s FDI in a more specific manner, the following unrestricted error correction model (UECM) of the ARDL model is estimated:

(6)

(7)

Where the εt in equation 1and 2 is the disturbance for the ARDL model. The null hypothesis that test the long run relationship of the model is β11=β12=β13=β14=0 and β21=β22=β23=β24=0 for equation 6 and 7 respectively, it specifies that there is no long run relationship. The alternative hypothesis contradicts with the null hypothesis as it states that at least one βj (j=11, 12, 13, 14) and βi (i=21, 22, 23, 24) is not equal to zero, it means at least one variable has long run relationship. If the computed F-statistic (Wald test) of ARDL bound test is greater than the upper bound critical value, we can reject the null hypothesis and conclude that the model has long run relationship. However, if the F-statistic is lower than the lower bound critical value, we cannot reject the null hypothesis and we can conclude that there is no cointegration in the model. Another possible outcome is that, if the value of F-statistic is recline between the lower bound and upper bound value we can just conclude that it is inconclusive (Narayan & Narayan, 2004).

3.3.2 Granger causality test

In order to know the short-run relationship between China’s foreign direct investment and economic growth in Malaysia, we apply the Granger causality in examining the relationship. In particular, we want to test whether the independent variable causes the dependent variable or vice versa or whether there is no causality at all or not. By applying the Granger causality test, we can automatically know that how well the previous dependent variable will explain the current dependent variable, and study the improvement of dependent variable’s expiration when introduced lagged in the model. They are three types of causal relationship in the Granger causality: (i) the dependent variable (Y) Granger causes the independent variable (X), (ii) the independent variable Granger causes the dependent variable, (iii) and no causal relationship at all.

Granger causality was described by Granger (1969) and a minor modification was done by Sims (1972). The Granger causality test means only a correlation between the current value of one variable and the past value of others; however it does not mean the movements of one variable cause the movement of another. Granger causality test helps us to answer the common question that "Does X causes Y"? There are several outputs that will be produced by the Granger causality test (Brooks, 1995):

If X causes Y, lags of X should be significant in the equation of Y, no vice versa. This means X Granger causes Y, a unidirectional causality runs from X to Y.

If Y causes X, lags of Y should be significant in the equation of X. It would be said that there was a ‘bi-directional feedback’ or ‘bi-directional causality’ if both sets of lags were significant.

If X is found to be Granger caused Y, but not vice versa, it would be said that variable X is strongly exogenous in equation Y.

It would be said that the variable X and Y is independent if there is no set of the lags are statistically significant in the equation for the other variable.

The direct way to test the Granger causality is to use the standard F-test of the restriction if the entire variable in the VAR are stationary:

β21(1)=β21(2)=β21(3)=….=β21(p)=0

Granger causality test is different from the other econometric tests because it assumes that all the variables are endogenous, therefore the researchers do not need to identify whether which variable is endogenous or exogenous. It is not the same as to test for exogeneity; it required that Yt not be affected by the contemporaneous term of Xt if Yt is to be exogenous. Nonetheless, Granger causality involves only the previous values effects of Xt on the current values of Yt. It therefore examines whether future values of Yt can be forecasted by using current and previous value of Xt.

We estimate the hypothesis by forming the MGDP as Y1 and CFDIM as Y2, and the other independent variables are labeled as CFDI, MOPEN, and FD for Y1; while MMS, EXR, and HCD for Y2. The εt and zt are the uncorrelated error term with white noise.

First of all, ΔX and ΔY are stationary time series, and εt and zt are uncorrelated white noise error term. Without including the lagged Y variables, we regress X on all lagged X terms and other variables and restricted residual sum of square (RSSR) is figured out. Next, we include the lagged Y terms and run the regression and the unrestricted residual sum of square (RSSUR) is figured out. Then, the null hypothesis in our research is lagged Y do not belong to the regression, that is H0: ∑βi=0; while the alternative hypothesis represents lagged Y does belong to the regression, that is H1: ∑βi≠0.

Following that, we use F-test (which follow the F-distribution) to perform the hypothesis testing, which is denoted as:

F= (RSSUR-RSSR)/m

RSSUR/ (n-k)

Where RSSUR indicates unrestricted sum of squared residual; RSSR represents restricted sum of squared residual; m is the number of lags; k represents the number of coefficient involved in the unrestricted regression. As a decision rule, if F-statistic greater than the critical value at a specific level, we can reject the null hypothesis, otherwise do not reject it. For the former case, we can conclude that the lagged Y terms belong to the regression. In other words, Y causes X. Finally, by including lagged X terns, we can repeat the whole process. In the other words, we interested to know whether the X Granger causes Y in the model or not.

3.4 Theoretical Model

3.4.1The Relationship between Trade Openness and Economic Growth

Trade can be decomposed into exports and imports. The nexus between exports and economic growth has been attributed to the potential positive externalities derived from exposure to foreign markets. In particular, there are three channels for which exports can spur growth (Awokuse, 2008). First, the expansion of export may accelerate output growth directly as a component of aggregate output. Second, the growth in exports can also indirectly influence the growth via different ways like stimulation of technological improvement, exploitation of economies of scale, greater utilization of capacity and resource allocation efficiency (Helpman & Krugman, 1985). Exports growth allows firms to take advantage of economies of scale that are external to firms in the non-export sector but internal to the overall economy. Third, export growth contributes in providing foreign exchange reserves that can be used to import raw material and intermediate goods, which will in turn enhance capital formation, and spurs output growth (Esfahani, 1991).

On the other hand, import growth which may serve as a complement in promoting economic performance as a whole, as compared to expanded exports (Awokuse, 2008). It is reasonable to hypothesize that the impact of imports on economic growth may be varying from that of exports. Besides, the technological transfer from industrial nations to less-developed countries through imports may serve as a crucial engine of growth. This is supported by the endogenous growth models which postulate that import can be a channel for long-run economic growth due to the accessibility to foreign knowledge and technology it provides to local firms (Coe & Helpman, 1995). Imports can be considered as sources of technology-intensive intermediate inputs of production (Mazumdar, 2001). As compared to exports, imports may play a more important role on economic growth because it is a medium of technology transfer.

In addition, we know that international trade may influent firm’s productivity through many different ways. In particular, via exports to developed nations (Clerides et al., 1998) and via imports of capital equipment and intermediate products (Markusen, 1989), it is a first channel of technology transfer. The geographic destinations of trade floes for both cases are very crucial. Firms importing capital and intermediate inputs from more sophisticated market must meet strict technical requirements to employ the sophisticated western technology. Similarly, firms exporting to advanced markets can learn more due to stiff competition, as well as to higher quality, to technical, safety and other standard requirements. Therefore, a higher propensity to trade with more developed countries may lead to a higher productivity level and faster total factor productivity (TFP) growth (Damijan et al., 2009).

With respect to economic literature, the impact of trade openness on economic growth is controversial. On one hand, higher proportion of exports and imports in GDP indicates greater trade volume and is expected to be positively related to economic growth (Hong & St. Juliana, 2010). This is supported by various researchers who found that trade openness positively affects GDP growth and they have statistically significant relationship (See: Edwards, 1998; Frankel & Romer, 1999; Winters, 2004; Wattanakul, 2009; Hong & Juliana, 2010; Constant & Yue, 2010). For example, according to Edwards (1998), more open economies have in fact experienced more rapid economic growth.The following variable is MOPENNESS defined as Malaysia’s trade openness and we believe that export and import will bring significant effect on the nation economic growth.

While several of these studies have documented empirical evidence reinforcing the existence of a long-run relationship between trade openness and economic growth, some others have rejected the trade openness-led growth hypothesis. In other words, they found an extremely contradictory result, in which trade openness negatively influent economic growth (See: Butkiewicz & Yanikkaya, 2010; Sarkar, 2008). For instance, according to Butkiewicz and Yanikkaya (2010), in mineral dependent economies, lower growth is resulted from trade openness because the development of domestic production is replaced by manufacturing imports.

The relationship between trade openness and economic growth may also be depending on situation. Honkapohja and Turunen-Red (2002); Madsen (2009); as well as Rao and Rao (2009)’s research support this argument. For example, according to Madsen (2009), openness does not, by and large, affect growth. But once we allow for the connection between openness and foreign knowledge, openness positively affects growth.

3.4.2 The Relationship between Financial Development and Economic Growth

Most of the theoretical literature regarding financial development and economic growth assume four different effects of financial transaction and development on economic performance as a whole. One of them is an informational effect, in which prior information about potential investment and capital are available (Levine, 2004). In addition, there is a risk management effect whereby the financial system contributes in diversifying risk (especially liquidity risk), and thus allowing for the financing of riskier but more productive investments and innovations (Bencivenga & Smith, 1991). Another effect is a volume effect and allocation effect, according to which financial transaction raises resources that can be channeled into investment while enhancing the allocation of resources devoted to investment. The last effect related to the provision of a cheap and reliable means of payment (Maswana, 2009).

In particular, the endogenous growth literature gives ample evidence that financial development affects economic growth and their relationship is statistically significant. According to theory, these two variables are linked in such a way that a well-developed financial system play a few major roles to improve the efficiency of intermediation by decreasing monitoring, information, and transaction costs, and thus this will in turn spur economic growth (Maswana, 2009). Obviously, it implies that without a sound financial system, economic growth seldom exist (Levine et al., 2000). Similarly speaking, there is no sustainable economic growth and no efficient financial depth if the financial system distorts the funds allocation under the circumstances of financial repression. Recent endogenous growth literature reinforced the significance of financial intermediaries in stimulating economic growth. Financial development can result in increasing return to capital, and a rise in the long-run growth rate via financial intermediation. As a result, financial development can have both the level effects and growth effects within the endogenous growth framework (Zhicheng, 2005).

With respect to economic literature, the impact of financial development on economic growth is controversial. On one hand, receiving considerable empirical support in previous studies is the positive relationship between financial development and economic growth. In other words, an improvement in financial development will spur growth. (See: Lensink, 2001; Calderon & Liu, 2003; Zhicheng, 2005; Liu & Hsu, 2006; Jun et al., 2007; Kar et al., 2008; Lee & Chang, 2009; Chee & Nair, 2010; Leitao 2010). For example, Kar et al. (2008) applied Johansen cointegration econometric method, found that there was a positive contribution from financial development to economic growth.

In contrast, some others have rejected the finance-led growth hypothesis. In other words, they found a contradictory result, in which financial development does not influent economic growth (See: Liang & Teng, 2006; Chimobi, 2010). For instance, Chimobi (2010) employed Johansen multivariate approach, found that there was no cointegration connection between growth and financial development (money supply, direct credit and private credit).

There were some other researchers produced rather mixed evidence on the role of financial development as a determinant of economic growth. In particular, the relationship between financial development and economic growth is depending on situation (See: Kar et al., 2010; Soukhakian. B, 2007; Soukhakian. N, 2007). For example, Kar et al. (2010) found that there is no clear cut on the direction of causality between financial development and economic growth for all measurements, and it is also found that the findings are country specific.

The financial development variable is used to gauge the resilient and sound financial system, which will in turn effective in promoting a country’s economic performance. If there is a positive and significant coefficient of this variable, we can conclude that financial development is effective in promoting economic growth. However, the relationship between financial development and economic growth remained controversial despite the fact that it had been studied by various researchers. Specifically, there were mix results being produced by various past authors. This phenomenon gives us a motivation to further analyze and confidently examine the correlation between the variables, with the hope that a more robust result can be produced.

3.4.3 Location Specific Model

In the China’s foreign direct investment (CFDIM) model, the exogenous variables that we study on are Malaysia’s exchange rate (EXR), market size (MMS) and human capital development (HCD).

Equation 7 showed an empirical model for the CFDIM (which is the dependent variable in the model); while equation 6 explains the relationship between China’s FDI and Malaysia’s economic growth, in which the coefficient of the former should be statistically significant.

In addition, for a small open economy like Malaysia, which is sensitively affected by the changes in the world price, the fluctuation of the exchange rate is especially important (Choong et al., 2005). In other words, exchange rate can serve as a tool to for adjusting the effects of such external shocks. We believe that an appreciation of the real exchange rate will decrease foreign direct investment (FDI). In other words, a devaluation of currency value will cause more FDI because if the Malaysian ringgit depreciates, this will enhance the competitiveness of the domestic commodities, and thus it will attract FDI and promote export (Choong et al., 2005). However, Henriques and Sadorsky (1996) argued that there is a positive correlation between exchange rate and FDI by providing example that with a lower Canadian dollar, Canada has had a superior economic performance compared with other advanced country like United State. The exchange rate (EXR) variable is the most concern element by multinationals firm investing in the host country. Most of the previous studies suggested that exchange rate has negative relationship with FDI, which means that an appreciation in the exchange rate will decrease the FDI flows.

The following is the MMS, or GDP per capita, means an approximation of the value of good produced per person in the country, equal to the country's GDP divided by the total number of people in the country. So, there should be a positive relationship between MMS and FDI. In other words, if Malaysia MMS is higher, the FDI inflows should be higher as well. The hypothesis of this variable is: the greater the MMS, the greater the business opportunity in Malaysia which will attract more FDI to the host country. We expect that the direction of this variable will be positive and it will significantly attract China’s FDI.

Finally is the HCD, in Malaysia’s development allocation for education and training has increased in 8th Malaysia plan compare with the 7th Malaysia Plan. Based on the research done by Awang (2004), improvement in human capital development (HCD) has become one of the attractions of FDI inflow to Malaysia. HCD should have a positive relationship with FDI inflow and it has been proven by most of the previous researchers, but still there is a small amount of researcher found out that HCD is not significantly affect FDI inflow even though it has positive relationship with FDI inflow. HCD consists of talent, skilled, education level, productive and managerial know-how workforce and has be concerned among the nations. We believe that the variable’s expected sign is positive and it will significantly affect China’s FDI (consistent with the theory employed by previous studies).