Which Type of Investments is Preferred to Achieve A Higher Economic Growth? A Case of the Upper-Middle-Income Economies

Type of Investments Achieve A Abstract Industrial revolution 4.0 requires the upper-middle-income economies to invest in critical areas such as technological advancement, infrastructure, internet of things, research and development, and so on. However, as majority of these economies are falling into the middle-income trap, they need huge supports from domestic and foreign investors to supply capital for growth stimulation. The issue on which type of investment should the governments rely on is crucial as it might help the countries to move out from the middle-income trap position. By using system GMM on four different growth models, it was found that gross saving is the main contributor to the economic growth of the upper-middle-income economies. Rather than domestic and foreign investments, the governments should accumulate more savings for future growth and development, which can be used as a source of capital especially in the areas of human capital development, technology, research, Internet of Things, in-line with the needs of industrial revolution 4.0.


Introduction
In-line with the industrial revolution 4.0, higher amounts of public and private investments are required to develop new infrastructures, enhance the performance of existing industries and train existing labors in meeting the industrial demand. For sure, it requires countries to inject higher amount of funds for the agenda of economic transformation. The sources of funds may come from the government itself, in the forms of national savings and public debt. The government can also acquire funds by attracting local and foreign investors to invest in critical areas such as education, infrastructure, research and development as well as technological advancement. These types of investments are highly crucial to achieve competitive advantage, thus allowing the countries to achieve economies of scale and penetrate bigger market size through export and import.
Nevertheless, attracting new investors to invest in a country is not an easy task. The macroeconomic foundation must be stable and resilient enough to confront with any shocks. The investors would be interested to invest in countries with high economic growth since it gives them confident that their return on investments will always be positive and significant. However, previous statistics show that some countries under the upper-middle-income economies (UMIE) were unable to maintain high economic performance, leading to demotion of their status from UMIE to lower-middle-income economies (LMIE). By definition, countries classified under the UMIE receive gross national income (GNI) per capita ranging from $3,896 to $12,055, while the countries with LMIE status earned GNI per capita from $996 to $3,895 (The World Bank, 2018). As illustrated in Table 1.1, 13 out of 56 countries within the UMIE have experienced demotion to LMIE since 1990 to 2019. The years of demotion varies from one year (Albania, Jordan, Tonga and Turkey) to fifteen years (Romania). Indirectly, it shows that the countries within the UMIE cannot use the same strategies to boost the level of economic growth as what they did when they were at the lower level of economic development. Applying the same strategies will lead to economic stagnation, thus leading to the middle-income trap (Bulman, Eden, & Nguyen, 2017). Focusing too much on existing products without any invention and innovation will lead to economic stagnation as other countries started to produce similar products with better quality. It if happens, the countries may no longer retain their competitive advantage on the products, thus limiting their ability to grow further. In mitigating the problem, the needs of support from domestic and foreign investors are highly crucial for economic transformation agenda for all UMIE including Malaysia. The funds injected by these investors can be invested in critical areas parallel to the development of the IR4.0.
Therefore, this paper is meant to investigate which type of investment is the most crucial investment for UMIE to achieve higher economic growth. This issue is worth to be investigated since the investment is regarded as one of the critical success factors for the economic growth and development. The structure of this paper is as follows. The next section reviews previous studies on how investment affects economic growth. The reviews cover both theoretical and empirical aspects of investment on growth. Follow suit is the third chapter that explains the research design, research method and model specifications as well as variables used in the analyses. The fourth chapter discusses the empirical results and findings while the final chapter explains the conclusions and policy recommendations.

Literature Review
The Solow growth model which is built from the neoclassical aggregate production function (Solow, 1956), stress three main causes of economic growth namely capital, labor and technology. These three resources are complementary to each other. A labor is able to produce more outputs if he is assisted by additional capital. However, any additional capital injected in the production process will have a diminishing return to the labor productivity (Snowdon & Vane, 2005). Therefore, the usage of capital in the production process will contribute the most to the economic growth when the capital itself is relatively scarce. In opposite, if the capital is relatively abundant in the economy, its impact on the economic growth may not be at the optimum level due to the diminishing return.
In an open economy, capital mobility across countries is possible. The countries with relatively abundant capital will be attracted to invest in other countries that give higher returns, thus accelerating the process of capital accumulation in the countries that have relatively scarce capital. Ultimately, countries with scarce capital can still achieve higher economic growth via the role of capital mobility.
Similarly, the endogenous growth model which was developed to overcome the weaknesses of the Solow growth model, agreed on the importance of capital to stimulate economic growth. Higher capital leads to better technological improvement. In the Solow growth model, Solow (1956) believes that the technological improvement is exogenous across countries. In other words, all countries have the same level of technological improvement as it is publicly available for free. Nevertheless, in the endogenous growth model, the technological improvement is considered as endogenous (Lucas, 1988;Romer, 1986). Each country has different level of technological improvement due to different abilities to absorb the knowledge. Thus, it is said that higher capital leads to improvement in technology and production of knowledge, which ultimately helps in stimulating the economic growth of a country. This is the main contribution of the endogenous growth model, which also highlights the importance of human capital in the growth literature. In this case, better human capital is developed due to the technological improvement which comes from the utilization of capital.
Empirically, a lot of research have been done on how capital investment affects the economic growth. The capital investment comes in various forms such as national savings (Mencinger, Verbic, & Aristovnik, 2015;Sulikova, Djukic, Gazda, Horvath, & Kulhanek, 2015), domestic investment ( (Rana & Wahid, 2017) and system GMM (Kim, Ha, & Kim, 2017), the results confirmed the positive effect of domestic capital investment on economic growth.
With regards to FDI as a source of capital, it gives positive multiplier effect to the economic growth in the cases of Malaysia (Alzaidy et. al., 2017), Korea (Kim & Pang, 2008) and Pakistan (Shahbaz & Rahman, 2012). It is mainly due to the knowledge spillover from the multinational companies to local businesses that indirectly support the domestic economic growth. By combining both arguments from domestic and foreign investments, both types of investment generate higher capital that helps to increase purchasing power of people via new employment opportunities. Higher consumptions and investment would then increase the economic growth of the countries.
Even though vast studies have found positive relationship between capital and economic growth, there were also studies that found otherwise. For instance, Thailand's domestic investment was found to adversely affect the economic growth from 1975 to 2018. The reason being is because of the ineffectiveness of the domestic investment on projects and activities that give lesser positive impacts to the economy (Raza, Aldeehani & Alshebami, 2020). By looking at previous literature, most of the studies were conducted on one specific country or a group of economies. Besides, previous research looked at the positive or negative effect brought by each different type of capital investment. Nevertheless, lack of studies has been conducted on how each type of these capital investments affect economic growth, specifically for the UMIE. Which one should the government focus on to strive for higher economic growth? This paper is meant to contribute to the existing literature by covering this loophole.

Methodology
The foundation of the growth model is based on the Cobb-Douglas production function. It can be written as follows: where Y is the aggregate output, K is the capital, L is the labor while α and 1-α are both representing the share of capital and labor in the national income. In this paper, we use three types of capital as previously discussed in past literature namely national savings (SAV), domestic investment (INV2) and foreign direct investment (FDI). Apart from K and L, other variables are also included in the growth model namely human capital (HC), trade openness (TO) and inflation (INF). Firstly, HC is included consistent with the endogenous growth model that highlights its importance. Secondly, TO is also included to represent the open economy that the countries within the UMIE are practicing. Thirdly, INF is added to indicate the macroeconomic stability of each country under investigation. Finally, the log of initial real GDP per capita (lnYi0) is included to illustrate how countries within the UMIE converge to the steady state of the equilibrium level (Islam, 1995). By combining all variables, the baseline model specification is written as follows: lnYit = αi + β1lnGDPPCi,t + β2lnLit + β3lnKit + β4lnHCit + β5lnTOit + β6lnINFit + εit (2) where Y is the GDP growth representing, i is cross section, t is time, α is a constant term, β is the coefficient for each variable and ε is the error term. The variable under investigation is K (capital). It represents three types of capital namely SAV, INV and FDI. Meanwhile, the remaining variables (L, HC, TO and INF) are regarded as the control variables.

Description and sources of the data
The description and the sources of the data are explained in Table 2.  Table. It is calculated based on the years of schooling and returns to education (Feenstra, Inklaar, & Timmer, 2015).

Scope of Study
The scope of this study is covering on the countries within the UMIE only since majority of these countries are in the middle-income trap. Out of 56 countries in the UMIE, only 32 countries were selected due to data availability. The list of countries under investigation is shown in the Appendix 1. The period of study is from 1990 to 2017. Since the impact of investment on growth can only be realized after few years, the data were averaged into nonoverlapping five-year period, leading to six time period. Following previous research (Karadam, 2018), the averaging procedure is conducted for two reasons. Firstly, we want to look at the long-run effects since the effects of investment cannot be realized on the same year. Secondly, the averaging procedure is initiated to avoid any structural break in the data that might influence the estimation results.

Method of Analyses
The first step in analyzing the data is by conducting the correlation analysis. The idea is to ensure all independent variables are free from multicollinearity problem. The next step is to remove any outliers in the models in order to ensure accuracy of the estimation results. To do so, Cook's D test is conducted to all four models. The test is done by calculating the cutoff distance (divide 4 with the number of observations). Since the number of observations is 189, any observations with cutoff distance more than 0.021 are considered as outliers. These outliers will be removed to ensure better accuracy of the results.
Once the outliers are removed, the selection of the right method for panel data estimation should be based on the number of time and cross sections. Since we have large cross sections (32 countries) but small time period (t=6), the appropriate method is the generalized method of moments (GMM). Apart from that, this method is also suitable as it is able to manage endogeneity issue in the set of the independent variables (Arellano & Bond, 1991). The endogeneity issue exists for our model because some of the independent variables can be the dependent variable at the same time. For instance, higher economic growth is one of the factors that can increase the domestic and foreign investors in the countries. In this regard, INV2 and FDI can be dependent variables against GDPG. Following previous growth literature, all variables in the growth models are treated as endogenous variables (Cieślik & Goczek, 2018). This issue cannot be tackled by using traditional panel estimators such as pooled ordinary least square, fixed effects and random effects model (Zhang, Hao, Lu, & Deng, 2018). Instead, it can be solved using GMM estimation method by adding lagged levels of regressors as the instrumental variables. In between system GMM and difference GMM, system GMM is more preferable since it is able to reduce biases and provide better estimation results (Blundell & Bond, 1998). Hence, this study uses system GMM to gather the estimation results.
The results from system GMM will only be valid if two conditions are met. Firstly, there should be no serial correlation in the error terms at the second order. It is tested by using Arellano-Bond test, with a null hypothesis of no serial correlation in the error terms at second order. Secondly, the instruments should be exogenous (Hansen, 1982). In this case, Hansen J test is conducted to ensure that the null hypothesis is not rejected (overidentifying restrictions are valid).

Empirical Results and Discussion
The results of the descriptive statistics for all variables used in this paper are tabulated in Table 3. As illustrated in  Apart from that, the correlation analysis as tabulated in Table 4 show no multicollinearity problem exists since the correlation values between all independent variables did not exceed 0.8. Thus, all independent variables can be included in the models.  Table 5 illustrates the results of the two-step system GMM for all four models. The results are highly consistent for all four models, indicating the robustness of the findings. Firstly, the LGDPPC as the convergence variable, meet the expected negative sign (Barro, 1991;Barro & Sala-i-Martin, 2004) with coefficients ranging from -1.626 to -3.485. Secondly, all investment variables have positive relationship with GDPG in all four models (Fashina et al., 2018;Pegkas, 2018), with FDI as the highest contributor to the economic growth. Thirdly, the signs of all control variables are parallel with previous literature, with negative sign for L (Kharusi & Mbah, 2018) and INF (Arčabić, Tica, Lee, & Sonora, 2018), and positive sign for HC (Karadam, 2018) and TO (Fashina et al., 2018). Finally, the p-values for both AR(2) and Hansen test suggest that the model is correctly specified and the instruments are valid.

Main Results
Besides, the most important source of growth for the UMIE is human capital. In all four models (1.1 to 1.4), the coefficients of the HC are the largest (excluding the convergence variable), ranging from 0.999 (in model 1.2) to 1.484 (in model 1.3). It indicates that the increase in the years of schooling generate higher returns to education and create pool of talents that can contribute to the economic growth in a long period of time (Ali, Egbetokun, & Memon, 2018). No doubt that education creates innovative mindset and forms a better quality of human capital and entrepreneurs. Within the four models, the coefficient of the HC is the highest when FDI is included in the model (model 1.4). Thus, it gives strong signal on the importance of FDI in accumulating capital, building human capital and achieving higher economic growth.
Before discussing the roles of investments on the economic growth, it is worth to look at the convergence variable denotes by LGDPPC. Consistent with the conditional convergence hypothesis, the coefficients of this variable are negative and significant in all four models (Barro & Sala-i-Martin, 2004;Karadam, 2018). The convergence hypothesis claims that the developing countries are able to grow faster than the developed economies due to diffusion of technology and lower costs of product imitation. In this case, the UMIE are able to converge to the steady state of the development path at a high growth rate relative to the developed countries. However, as the countries converge to their steady state of the development path, other countries will start to catch-up with the development via product imitation and technological diffusion that can be obtained mainly from the FDI.   1. ***, ** and * indicate 1%, 5% and 10% significant level respectively. 2. Values in parentheses are the standard errors.

Discussion on lnSAV to Economic Growth
In between all types of capital investments, lnSAV is the highest contributor to the economic growth due to its largest coefficient relative to lnINV2 and lnFDI. Parallel to previous studies (Gómez-Puig & Sosvilla-Rivero, 2018a), similar positive finding can be seen from the gross savings with the coefficient value of 0.2471 (model 1.1) and 0.2237 (model 1.4). It implies the deepening of the financial development in the UMIE, which assist the countries to increase their wealth and accumulate higher savings for future development (Grigoli, Herman, & Schmidt-Hebbel, 2018). Moreover, a significant positive relationship between lnSAV and lnY illustrates that the countries are channelling their savings for investment in productive purposes, thus boosting the economic growth (Wan Azman Saini, 2009).
Apart from that, gross savings lead to higher economic growth especially when the economy is facing with lower macroeconomic instability (Grigoli et al., 2018). In order to prove this statement, we should look at the coefficient of the lnINF since it measures the macroeconomic instability. By observing the models with lnSAV (model 1.1 and 1.4), lnINF has the lowest coefficients in both models with -0.0554 and -0.0696 respectively. Even though the signs of lnINF in both models are negative, the coefficient is very low. It indicates that the macroeconomic instability in the UMIE were under control. It gives good sign for the private sector to consume and invest more, thus playing the role as the engine of growth. The private sector's contribution helps the government to save more and invest the funds in productive expenditures, then contributing to the national growth and development.
If the gross savings are used for unproductive or corrupted purposes, it might not help in increasing the economic growth of a country. It is proven statistically, in which Gabon has the highest percentage of average gross savings to GDP from 1990 to 2019 (49.06 percent). Nevertheless, in 2018, Gabon was considered as among the corrupted countries (Transparency International, 2018) since the corruption perception index was close to 0 (scored at 31 out of 100). As a note, the score is ranging from 0 (highly corrupted country) and 100 (the cleanest country in terms of corruption). Due to misallocation or mismanagement of funds particularly from gross savings, the average GDP growth per capita was at the lowest level (-0.568 percent) as compared to the other UMIE (refer Error! Reference source not found.). However, when combining all other UMIE into the panel data analysis, lnSAV still gives positive effects to lnY despite a special case for Gabon.

Discussion on lnINV2 to Economic Growth
As a measure of domestic investment, lnINV2 is proven to have positive relationship with lnY. It is shown in the coefficients of lnINV2 in model 1.2 and 1.4, with 0.2092 and 0.2030, respectively. By comparing model 1.1 and model 1.2, lnINV2 is having a similar role relative to lnSAV in influencing lnY due to similar coefficient (0.2). One possible reason is due to the ability of the domestic investment in generating new employment opportunities to the public. An increase in the number of jobs will not just increase the purchasing power of the people. It is also helpful in reducing the poverty and income inequality, as well as increasing the growth of the country via higher private consumptions.
Moreover, the positive relationship postulates higher domestic investment have been channelled to research and development (R&D), technological progress and other productive purposes, which in turn leads to an increase in the productivity growth of the countries (Dao, 2018). Higher productivity will motivate the private sector to produce more goods and services at a lower cost, thus increasing the demand for labour. As long as the investment is supported by technological progress, UMIE will be capable in creating new knowledge and technology by combining both capital and skilled labours efficiently. This, in turn, will gives value added to the economic growth of the UMIE.
Nonetheless, the coefficient of lnINV2 in model 1.4 (0.2030) is lower than the coefficient of the same variable in model 1.2 (0.2092). It indicates that the contribution of the domestic investment is declining when the UMIE is also relying on the foreign investment as a source of growth. It might be due to fierce competition that domestic investors have to embrace as the foreign investors are normally equipped with far better technological progress (Krstevska & Petrovska, 2012). Even so, this competition is good for the economic growth since the domestic and foreign investors can learn from each other, sharing expertise and transferring knowledge to ensure better growth.

Discussion on lnFDI to Economic Growth
The result for lnFDI is in agreement with our expectation. In achieving high economic growth, the UMIE needs to attract more investments from abroad, especially in relation to hightechnology, digitalization and so on. High inflow of FDI would in turn, leads to a better human capital development due to transfer of skills, knowledge and technology that are deemed essential for economic transformation. The spill over from the FDI is transferred into the human capital investment in the forms of training, thus further helps the UMIE to achieve higher economic growth.
Consistent with previous research, the positive effect of lnFDI is partly contributed by the level of trade openness of a country (Akram, 2016;Jayasuriya, 2011). It can be shown by looking at the significant coefficient of the lnTO in the third model, amounting to 0.2432. As countries within the UMIE are willing to involve more in trade and partnership with the rest of the countries, the trade activities help them to receive valuable technology and expertise from abroad via the roles of FDI, thus boosting the productivity of the country.

Conclusion
The most relevant conclusions are as follows. Firstly, the economic growth of the UMIE is positively depends on both domestic (lnSAV and lnINV2) and foreign investments (lnFDI). When combining all the three investment variables into a model (model 1.4), lnSAV is the most contributing factor in achieving high economic growth, followed by lnINV2 and lnFDI.
Secondly, since human capital is proven to be the most important factor influencing the economic growth (in model 1.1 and 1.2), the countries within the UMIE should invest more on the human capital to ensure their labours are able to generate new ideas, knowledge and technology that are essential for future economic growth. Besides, high reliance on the foreign investors might be the right decision if they want to achieve a high-income nation. It is because FDI assists in human capital development through the transfer of knowledge, expertise as well as technological advancement. In this case, the governments should attract more FDI into the countries by introducing or enhancing the current tax incentives. Besides, the institutional and business environments' quality should be preserved to ensure rapid inflows of FDI to the countries. This is important as the foreign capital, expertise and technological advancement from other countries can help the UMIE to move out from the middle-income trap, hence uplift their status into HIE while implementing the IR4.0 successfully. Sole reliance on domestic investment and domestic savings are insufficient for that matter. This paper contributes empirically to the body of knowledge by integrating all three types of investment into the endogenous growth model. Theoretically, as all of these investments are positively contributed to the economic growth and development, their contribution might vary for a country or a group of economies. For instance, this research found that the FDI is not the most significant contributor to the economic growth for the UMIE even though it gives positive multiplier effects to the countries in the forms of technological and knowledge