The Determinants and Growth Effects of Foreign Direct Investment: A Comparative Study

Abstract

This study examines the factors determining inward foreign direct investment (FDI) and its effects on productivity, ultimately contributing to economic growth. Using a two-step generalized method of moments (GMM) approach, we analyzed a panel of 84 countries, comprising 34 OECD and 50 non-OECD countries, from 2010 to 2019. The findings suggest that FDI positively impacts productivity and benefits both OECD and non-OECD countries. Economic freedom plays a significant role in attracting FDI, particularly in OECD countries, and contributes to economic growth in non-OECD countries. However, economic freedom alone does not guarantee strong economic growth in OECD countries but significantly enhances growth in non-OECD countries. The results also highlight that only economies with robust economic infrastructure and development levels benefit more from FDI. It appears that FDI by itself has no direct effect on output growth. Instead, the impact of FDI is contingent on the level of economic freedom in the host countries. This paper presents a key finding on how policy decisions influence the effects of foreign capital investment on productivity and income. It indicates that countries promoting economic freedom can more effectively leverage productivity gains from FDI.

1. Introduction

Economic openness attracts foreign direct investment (FDI), which not only brings in capital but also transfers knowledge and technology to host countries that would not otherwise have access to it. As investment from other countries links FDI host countries to the global market, global value chains provide opportunities for the countries to diversify their exports and expedite their integration into the global economy, as noted by Dollar (2017) in his Brookings feature article.1 Through global value chains, host countries are exposed to broader economic activities in widely separated locations that are interconnected. Local firms may benefit from substantial economies of scale, strengthening their competitive global market positions. The World Bank (2021) underscores the crucial role of policies that attract and facilitate FDI entry in enhancing investment promotion capacity. These policies lead to spillover effects through supply chain linkages, which bring new capital, technology, and knowledge into business practices, enabling local firms to better withstand the onslaught of global, scalable firms. Serpa and Krishnan (2018) indicate that firms interacting within a supply chain affect each other’s productivity, and productivity spillovers are more pronounced when the supply chain is more concentrated. Thus, countries that gain from productivity improvements accrued from FDI inflow are more likely to pursue policies favoring economic openness.

Productivity spillovers from multinational activities to local firms come with knowledge and technology to an extent that is non-excludable and non-rival in production. Blomström and Kokko (1998) identify three main channels where productivity spillovers occur from FDI inflows to host countries, including demonstration effects through the imitation of newer technologies, competition effects through competitive pressure, forcing local firms to improve their production efficiency, and knowledge spillover effects through skilled labor mobility. Developing countries tend to attract a sizeable share of FDI through comparative advantage mechanisms and more substantial capabilities to pursue economic growth. For these countries to achieve productivity spillover gains, FDI inflows need to be stable and complementary to the labor force structure within the local economy, where FDI inflows would bring the capital formation to an extent sufficient to absorb the growth in the labor force, adjusting labor and capital to the equilibrium where productivity spillover effects can be prevalent. Consequently, countries that better capture the effect of demonstration or contagion are more likely to benefit from FDI inflows.

One aspect of low wages in developing countries can be attributed to the lack of productive capital, which is the primary factor driving structural unemployment. For wages in these countries to rise, capital investments need to be sufficient to reach the steady state growth rate of labor supply to boost labor productivity, and FDI can be a source for capital investment to expand to the level where productivity will grow, rendering in terms of skill development and productivity spillover. Consequently, many developing countries turn to FDI attraction, seeking foreign capital to mitigate capital constraints and fill the resource gap in their quest for economic development. In their empirical studies, Lipsey and Sjöholm (2004), and Hijzen et al. (2013) put forth a view that multinational corporations (MNCs) tend to offer wage premia compared to their domestic counterparts with which the gap between wages in local firms and MNCs forces domestic firms to face competitive pressure in the labor markets for higher wages. MNCs share knowledge or technology with their suppliers, which they then implement to increase productivity. In return, MNCs would demand higher-quality goods and resources from suppliers, forcing local firms to be more productive. Thus, productivity spillovers occur as FDI creates benefits that local suppliers and other firms share in the local markets.

This study examines the potential of FDI inflows to promote economic growth across countries at various stages of development. We specifically investigate the ways in which FDI contributes to economic growth through productivity enhancements and assess the influence of economic freedom on a country’s ability to attract FDI. We emphasize that FDI’s effects on productivity depend positively on labor market concentration in the host country and attempt to measure the effects of productivity by exploring the ratio of capital to the stock of labor efficiency rather than focusing on horizontal or vertical productivity spillovers, as is commonly done. From a macroeconomic perspective, our approach to the FDI effect on productivity allows us to examine the overall impact of FDI inflows for an economy rather than within a specific sector or the level of the economy where the viewpoints would be relatively segmented.

We apply the system generalized method of moments (GMM) approach proposed by Arellano and Bover (1995), and Blundell and Bond (1998) to estimate the impact of FDI inflows on economic growth through productivity spillover effects. Based on a panel of 84 countries from 2010 to 2019, the findings suggest that FDI positively impacts productivity and benefits both OECD and non-OECD countries. Economic freedom is pivotal in attracting FDI inflows, particularly in OECD countries, and contributes to economic growth in non-OECD countries. While economic freedom does not guarantee robust economic growth in OECD countries, it significantly enhances growth in non-OECD countries. However, only economies with strong infrastructure and high levels of development benefit more from FDI. FDI by itself has no direct effect on output growth. Instead, the impact of FDI is contingent on the level of economic freedom in the host countries.

This paper explores the factors that influence inward FDI and its positive effects on productivity. While many policies aim to attract foreign investment, the impact of FDI inflows on economic growth varies significantly depending on the specific circumstances of each country. We argue that nations promoting greater economic freedom are more capable of harnessing the productivity benefits associated with FDI, ultimately enhancing their potential for sustained economic growth. Our study focuses on various groups of transition economies and OECD countries. A key conclusion of this paper is that policy decisions play a crucial role in shaping how foreign capital investment impacts productivity and income. Our findings suggest that countries encouraging economic freedom are more successful in leveraging the productivity gains from FDI.

The remaining sections of the paper are organized as follows. Section 2 provides an overview of the relevant literature. Section 3 describes the data and econometric methodology applied in the study. Section 4 presents a preliminary finding explaining why system GMM estimates are superior to difference GMM estimates in this study, followed by our empirical findings of the GMM estimates with interaction terms and data segmented into transition and OECD economies. The main findings are summarized with a concluding remark in the Section 5.

2. Literature Review

Despite their differing viewpoints, the neoclassical and endogenous growth models serve as foundational frameworks for much of the empirical research exploring the relationship between FDI and economic growth. The neoclassical growth model, particularly the Solow theory, posits that FDI enhances the capital stock and stimulates economic growth in the host country by promoting capital formation (Brems 1970). However, because of diminishing returns on capital, the influence of FDI on growth is comparable to that of domestic investment. In this framework, FDI plays a role in economic growth in the short term as countries transition toward a new steady state (Solow 1956).

Conversely, the endogenous growth theory asserts that FDI significantly drives economic growth through capital formation and technology transfer. As Romer (1994) noted, FDI enhances economic growth by facilitating technology transfer from developed nations to developing ones. Moreover, the training provided for the workforce and management improves knowledge and strengthens human capital. Consequently, the accumulation of human capital and technological advancements are vital factors influencing the spillover effects of FDI on the host country’s economic growth (De Mello 1997, 1999). Ultimately, the endogenous growth model suggests that FDI may be more productive than domestic investment due to FDI’s ability to integrate advanced technologies into the production processes of host economies. From this perspective, the technological spillovers associated with FDI can serve to alleviate the effects of diminishing returns to capital, thereby promoting sustained economic growth over the long term.2 FDI appears to play a crucial role in fostering the economic growth of host countries.

While theoretical frameworks suggest a direct link between FDI inflows and economic growth, empirical studies provide inconsistent evidence regarding the impact of FDI on productivity. Blomström et al. (2003) argue that the advantages of FDI spillovers can only be fully realized when local firms possess the capacity and motivation to leverage foreign technologies and skills effectively. They emphasize that FDI alone does not ensure the transfer of these foreign technologies and skills to domestic industries. Numerous studies have yielded varied results to clarify the relationship between FDI and economic growth, prompting a series of meta-analyses to examine the effects of productivity spillovers associated with FDI. One such meta-analysis identified evidence of publication bias and concluded that the type of data employed—whether derived from industrial or plant-level sources, representative of developing or developed countries, or more recent—does not significantly influence the findings (Havránek and Iršová 2010). Demena and van Bergeijk (2017) investigate 1450 estimates from 69 empirical studies published in 1986–2013 and note the inconsistency of reported FDI spillover findings caused by publication bias, characteristics of data, and model misspecification.

Meyer and Sinani (2009) suggest that the relationship between a country’s stage of economic development and the spillover effects of FDI appears to be curvilinear. Economies characterized by low and high levels of GDP per capita derive benefits from FDI spillovers; however, transition economies situated in the intermediate range exhibit a decreasing influence. Wooster and Diebel (2010) also present weak evidence of horizontal spillovers in developing countries, contrasting prior studies. However, research examining spillovers to domestic firms in the same sector is relatively limited. Iršová and Havránek (2013) find similar results to Wooster and Diebel (2010) with no evidence of horizontal spillovers. They argued that past levels of FDI do not affect current spillovers, implying a linear effect, wherein a country saturated with FDI may still reap spillover benefits.

Empirical research utilizing data from specific countries offers compelling evidence that supports the relationship between FDI and economic growth. For instance, Asheghian (2004) and Salehizadeh (2005) confirmed a positive and significant relationship between FDI and economic growth in the United States. Kornecki and Raghavan (2011) noted that countries in Central and Eastern Europe view FDI as a crucial tool for developing and modernizing their economies. Additionally, several studies have shown that FDI inflows have a strong and positive effect on economic growth in China (Tian et al. 2004). Empirical analyses by Fang and Liu (2007), Sharahili and Liu (2008), and Omer and Yao (2011) found a bi-directional causality and long-term relationships between inward FDI and economic growth in both China and Malaysia.

The extant body of literature mainly supports the idea that FDI positively impacts economic growth, with productivity spillovers primarily seen as beneficial to domestic firms through the transfer of technology and knowledge from MNCs. Borensztein et al. (1998) test the effect of FDI on economic growth in a framework of cross-country regression and find that FDI is an important vehicle for the transfer of technology, contributing to growth in a larger measure than domestic investment. However, their empirical results also suggest that FDI is more productive only when the host country has a minimum threshold stock of human capital. Li and Liu (2005) investigate whether FDI affects economic growth based on a panel of data for 84 countries over the period 1970–99. They find that FDI directly promotes economic growth in developing countries and also contributes indirectly through its interaction with human capital. However, FDI combined with a technology gap has a significantly negative impact.

Amann and Virmani (2014) examine the feedback effect of FDI on productivity and find that the effect of FDI on enhancing productivity growth is more significant for R&D-intensive developed countries investing in emerging economies. Gorodnichenko et al. (2014) present evidence to partially support Blomström and Kokko (1998), suggesting that domestic firms experience a faster increase in efficiency after supplying to MNCs. However, domestic firms purchasing from or competing with MNCs in the same industry do not appear to gain positive spillovers. This could be due to the cost of implementing new technology to compete with MNCs in the short run. Based on a sample of 17,625 Chinese manufacturing firms, Liu (2008) shows that FDI spillovers lower domestic firms’ productivity levels in the short run and that backward linkages are the most statistically significant channel for technology spillovers to occur, potentially due to the demand for high-quality goods supplied to MNCs by domestic firms. The result suggests that productivity spillovers from FDI are still positive and substantial overall.

Building on the channels by Blomström and Kokko (1998), Azman-Saini et al. (2010) examine 85 countries and generate evidence of productivity spillovers from FDI conditioning on high levels of economic freedom. They highlight how economic freedom and FDI jointly positively and significantly affect growth. However, FDI alone does not impact growth, concluding that this occurs because economic freedom facilitates FDI spillovers. Naanwaab and Diarrassouba (2016) also contend that economic freedom is a positive determinant of FDI but also note that this occurs only when economic freedom is sufficiently high, and the most critical components of economic freedom are legal structure and property rights, regulation, and freedom to trade. This is similar to Iamsiraroj and Ulubaşoğlu (2015) who find that financial development and trade openness are essential for a country to have the absorptive capacity for FDI to impact growth.3 However, the effect subsides once the economic development and trade openness reach certain levels. Sambharya and Rasheed (2015) posit that economic and political freedom affects countries’ institutional environments, which can influence a firm’s decisions to enter markets. Higher economic freedom overall leads to increased FDI inflows. However, only four out of five of the Heritage Foundation’s economic freedom index components influence FDI inflows. The exception of trade and investment freedom contrasts with Azman-Saini et al. (2010), who find that freedom to trade is essential.

Naanwaab and Diarrassouba (2016) show that economic freedom alone cannot attract FDI inflows in developing countries; human capital and economic freedom are jointly substantial. Human capital is critical in enhancing absorptive capacity so domestic firms can maximize productivity spillovers by utilizing foreign knowledge and technology (Blomström et al. 2003). (Balasubramanyam et al. 1999) also find that for human capital to act as a conduit for FDI to affect growth, a threshold of human capital is necessary. Similarly, Bengoa and Sanchez-Robles (2003) argue that FDI positively correlates with economic growth conditioned upon adequate levels of human capital. Fukase (2010) infers that FDI affects growth through human capital differently depending on the channel, industry, and institutions. Wooster and Diebel (2010) note that labor quality is inversely associated with spillover absorption, where higher labor quality leads to decreased productivity spillover benefits from FDI. Kristjánsdóttir (2010) implies that MNCs are less attracted by human capital and labor education levels as they originate from highly skilled labor countries. The labor cost difference between countries was a more significant determinant of FDI, as MNCs are more likely to take advantage of cost-saving labor.

Restrictive labor laws could inhibit knowledge transfer between MNCs and domestic firms through the labor movement. Using data from Vietnam and China, (Vu et al. 2008) analyze the impact of FDI inflows on growth through labor in different economic sectors. They find that FDI inflows significantly and positively affect growth through indirect interaction with labor in both countries. However, the effect is not equally distributed among sectors. High levels of regulation could also impede the efficient allocation of resources, giving less opportunity for domestic firms to acquire funding to adopt new technologies; strong protection of property rights would entice more technology-oriented FDI to develop their R&D domestically; and freedom to trade would allow MNCs to export easily while producing goods at a lower cost in developing countries, creating competition and relationships with domestic firms in the process.

3. Data and Methodology

This study focuses on countries in distinct stages of economic development. In parts of the analysis, the distinction is made between OECD and non-OECD countries. A panel of 84 countries includes 34 OECD countries and 50 non-OECD countries over the 2010–2019 period. The variables included in the analysis are real GDP per capita, FDI inflows, economic freedom index, gross capital formation, population growth rate, labor force participation rate, and labor productivity. FDI inflows are the net inflow as a percentage of real GDP. Real GDP and gross capital formation are in constant 2015 US dollars. Labor productivity is output per employed person in constant 2015 US dollars. The population growth rate is the ratio between the annual increase in the population size and the total population. The labor force participation rate is the percentage of a country’s working-age population actively engaged in the labor market. Economic freedom is an index measuring the degree of freedom in a country comprised of five major components: (1) size of government, (2) legal structure and security of property rights, (3) access to sound money, (4) exchange with foreigners, and (5) regulation of capital, labor, and business.

Real GDP and gross capital formation were drawn from the World Bank World Development Indicator. The data on FDI inflows were obtained from the United Nations Conference on Trade and Development data center (UNCTAD). Economic freedom is the index compiled by the Fraser Institute. Population growth rate is compiled by the World Health Organization (WHO). Labor productivity and labor force participation rates were obtained from the International Labor Organization Statistics Database (ILOSTAT).

Table 1. Summary statistics for the period 2010–2019.

Variable	Mean	Std. Dev.	Minimum	Maximum
Full Sample
GDP Growth (%)	3.14	2.66	−12.09	20.11
FDI (% of GDP)	5.46	18.70	−40.29	280.13
Economic Freedom	7.16	0.81	4.45	8.62
Capital Cumulation (% of GDP)	18.32	20.30	−263.56	71.16
Population Growth (%)	1.18	1.37	−1.319	12.25
Labor Force Participation Rate (%)	61.99	8.912	39.25	89.05
Labor Productivity	38,557.87	39,987.14	928.9	181,776
Non-OECD Countries
GDP Growth (%)	3.75	2.72	−12.09	10.19
FDI (% of GDP)	6.64	22.87	−24.90	280.13
Economic Freedom	6.78	0.80	4.45	8.25
Capital Cumulation (% of GDP)	18.32	24.87	−263.56	71.16
Population Growth	1.65	1.50	−0.92	12.25
Labor Force Participation Rate (%)	62.66	10.71	39.25	89.05
Labor Productivity	14,119.7	13,082.98	928.9	68,993
OECD Countries
GDP Growth (%)	2.22	2.29	−11.30	20.11
FDI (% of GDP)	3.73	9.52	−40.29	81.33
Economic Freedom	7.73	0.40	6.56	8.62
Capital Cumulation (% of GDP)	18.34	10.46	−55.55	66.08
Population Growth	0.50	0.74	−1.32	2.94
Labor Force Participation Rate (%)	61.00	5.12	47.97	77.75
Labor Productivity	74,496.36	39,103.87	14,338	181,776

reports summary statistics for GDP growth rate, FDI inflows, domestic capital formation in percentage of GDP, population growth rate, labor force participation rate, and labor productivity changes over the sample period based on the respective full sample, OECD, and non-OECD countries.4 Non-OECD countries appear to have higher GDP and population growth rates than OECD countries, while average labor productivity in the former is notably lower than in the latter. In terms of GDP size, transition economies also receive approximately twice the FDI inflows as a percentage of GDP than OECD economies.

Figure 1 displays the average annual real GDP growth rate, FDI inflows, and economic freedom for the sampled countries over the entire period (2010–2019). The fitted line in Figure 1a shows a weak positive relationship between FDI and economic growth (R² = 0.038). Meanwhile, the fitted line in Figure 1b illustrates a negative correlation between output growth and economic freedom (R² = 0.092). The economic freedom of high-growth transition economies appears to cluster around the lower end of the index, while FDI inflows contribute positively to economic growth. However, the correlations do not imply the causation of FDI to economic growth. If economic freedom is vital in fostering economic growth, one would expect that FDI inflows and economic freedom are complements.

Due to the dynamic nature of the regression equation, FDI may exhibit endogeneity in its relationship with economic growth, which can lead to simultaneity bias when utilizing country-specific dummy variables. To address this issue, this study employs the GMM panel estimator developed by Arellano and Bond (1991), Arellano and Bover (1995), and Blundell and Bond (1998). This approach effectively accounts for country-specific effects, which cannot be adequately addressed with country-specific dummies given the dynamic structure of the model.

The model specification is broadly consistent with other panel data approaches that use FDI to estimate economic growth (for instance, Caselli et al. 1996; Alfaro et al. 2004). Our model also includes labor productivity and the economic freedom index to measure the effect of productivity spillover and economic openness. Three conditional variables, gross capital formation, population growth rate, and labor force participation rate, are vital factors of production. The former measures the investment levels of countries, and both latter variables are used to gauge labor force growth in the countries. The economic freedom index will likely capture most other pertinent variables, such as trade openness and government policies.

We first specify a growth level equation as follows:

$$y_{it}-y_{it-1}={{\beta }_1{y}_{it-1}+\beta }_2{FDI}_{it}+{\beta }_3{EF}_{it}+{\beta }_4{CF}_{it}+{{\beta }_5{PG}_{it}+\beta }_6{LF}_{it}+{\beta }_7{LP}_{it}+{\eta }_i+{\gamma }_t+{\epsilon }_{it}$$

(1)

where y_it denotes the logarithm of real GDP in country i in year t; FDI_it is foreign direct investment, EF_it is economic freedom, CF_it is gross capital formation, PG_it is the population growth rate, LF_it is the labor force participation rate, and LP_it is labor productivity. η_i is an unobserved country-specific effect reflecting differences in the initial level of efficiency. The period-specific intercept γ_t captures real income changes that are common to all countries, and ε_it is the error term. Equation (1) can be rewritten as a dynamic model in the level of GDP by adding y_it−1 to both sides as follows.

$$y_{it}={{\alpha }_{1}y}_{it-1}+{\beta }_2{FDI}_{it}+{{\beta }_{3}EF}_{it}+{\beta }_4{CF}_{it}+{{{\beta }_5{PG}_{it}+\beta }_{6}LF}_{it}+{{\beta }_{7}LP}_{it}+{\eta }_i{+\gamma }_t+{\epsilon }_{it}$$

(2)

where α = (1 + β₁). Equation (2) explicates the dynamic structure of the growth equation. However, the OLS estimate of the coefficient on the lagged dependent variable could bias upward as $y_{it-1}$ is positively correlated with the country-specific effects η_i. As proposed by Arellano and Bond (1991), one approach to address this issue is to employ lagged levels of the regressors as instruments, given that the error term is not serially correlated, and the lagged explanatory variables are weakly exogenous. This alternative approach addresses the presence of the lagged endogenous variable by first differencing Equation (2) to eliminate country-specific effects as follows:

$$∆y_{it}={{\alpha }_1∆y}_{it-1}+{\beta }_2{∆FDI}_{it}+{{\beta }_3∆EF}_{it}+{\beta }_4{∆CF}_{it}+{{{\beta }_5{PG}_{it}+\beta }_6∆LF}_{it}+{{\beta }_7∆LP}_{it}+{∆\epsilon }_{it}$$

(3)

where the first difference equation removes omitted-variable bias and country-specific effects. Then, following Arellano and Bond (1991), we set the following moment conditions:

$$E\left[y_{it-s}∆{\epsilon }_{it}\right]=0 \mathrm{for} s\ge 2;t=3,\dots T$$

$$E\left[{FDI}_{it-s}∆{\epsilon }_{it}\right]=0 \mathrm{for} s\ge 2;t=3,\dots T$$

$$E\left[{EF}_{it-s}∆{\epsilon }_{it}\right]=0 \mathrm{for} s\ge 2;t=3,\dots T$$

$$E\left[{CF}_{it-s}∆{\epsilon }_{it}\right]=0 \mathrm{for} s\ge 2;t=3,\dots T$$

$$E\left[{PG}_{it-s}∆{\epsilon }_{it}\right]=0 \mathrm{for} s\ge 2;t=3,\dots T$$

$$E\left[{LF}_{it-s}∆{\epsilon }_{it}\right]=0 \mathrm{for} s\ge 2;t=3,\dots T$$

$$E\left[{LP}_{it-s}∆{\epsilon }_{it}\right]=0 \mathrm{for} s\ge 2;t=3,\dots T$$

Despite being able to control country-specific effects and simultaneity bias, difference GMM estimates potentially suffer from instrumental problems with small samples. If the dependent variable was persistent and close to being a random walk, applying difference GMM estimation would yield both a biased and inefficient estimate of α₁ in finite samples, particularly acute with a relatively shorter estimation period.

Moreover, economic growth is dynamic and dependent on past levels, and the independent variables are not strictly exogenous, given that FDI and domestic investment may be correlated with past and current realizations of the error. Empirical research in international economics acknowledges that FDI, a dynamic process strongly dependent on past observations, could also potentially lead to disturbances that are likely heteroskedastic and serially correlated. One source of endogeneity arises from the possibility that current values of FDI are a function of past economic performance. Blundell and Bond (1998) attribute the poor precision and significant finite sample bias of the difference GMM estimator to lagged levels of the independent variables being weak instruments and propose using system GMM to address the simultaneity bias in the independent variables. The system dynamic panel GMM estimator effectively avoids the deviation caused by this endogeneity problem by utilizing Equations (1) and (3) to provide more consistent estimates as a second condition for utilizing system GMM. The approach reduces the biases and imprecision by using additional moment conditions wherein difference equations use level instruments, and level equations use difference instruments. The estimated data, which include a short period and a large number of countries, meet the requirements of a small period and a large number of countries to be estimated by GMM empirically.

Following Arellano and Bover (1995), we include the additional moment conditions for the second part of the system, the level equation.

$$E\left[(y_{it-s}-y_{it-s-1})({\eta }_i+{\epsilon }_{it})\right]=0 \mathrm{for} s=1$$

$$E\left[({FDI}_{it-s}-{FDI}_{it-s-1})({\eta }_i+{\epsilon }_{it})\right]=0 \mathrm{for} s=1$$

$$E\left[({EF}_{it-s}-{EF}_{it-s-1})({\eta }_i+{\epsilon }_{it})\right]=0 \mathrm{for} s=1$$

$$E\left[({CF}_{it-s}-{CF}_{it-s-1})({\eta }_i+{\epsilon }_{it})\right]=0 \mathrm{for} s=1$$

$$E\left[(PG-{PG}_{it-s-1})({\eta }_i+{\epsilon }_{it})\right]=0 \mathrm{for} s=1$$

$$E\left[({LF}_{it-s}-{LF}_{it-s-1})({\eta }_i+{\epsilon }_{it})\right]=0 \mathrm{for} s=1$$

$$E\left[({LP}_{it-s}-{LP}_{it-s-1})({\eta }_i+{\epsilon }_{it})\right]=0 \mathrm{for} s=1$$

The GMM estimators are applied with two-step variants using the optimal weighting matrices, in which the moment conditions are weighted by a consistent estimate of their covariance matrix. This would be asymptotically more efficient, as explained in Arellano and Bond (1991). However, using the two-step estimator in small samples would potentially result in the proliferation of instruments, leading to biased standard errors and parameter estimates. To solve this problem, Windmeijer (2005) proposed a finite sample bias-corrected standard error formula for the two-step linear GMM. Following Roodman (2009), we use the above moment conditions and employ the two-step estimator.

We further explore the role of FDI in economic growth, focusing on productivity spillovers and economic freedom. To clarify these relationships, we introduce an interaction term to assess their contingency. If this term is positive and significant, it suggests that the impact of FDI on growth is enhanced by increased productivity and economic freedom. We anticipate that this interaction term will be more significant in transition economies compared to OECD countries. This is because the benefits of FDI inflows on economic growth are generally greater during the developing phase of an economy, especially as these countries implement more flexible economic policies. Strengthening the freedom of economic activities is crucial for facilitating FDI spillovers. This hypothesis is consistent with recent studies, which indicate that the recipient countries’ absorptive capacity influences the FDI’s effectiveness on growth. Thus, the advantages of economic freedom can be seen as both direct and indirect.

Following Equations (2) and (3), we interact FDI with respective labor productivity and economic freedom and use each interaction term as a regressor. To ensure that the interaction term does not proxy FDI or the level of labor productivity and economic freedom, both latter variables were included in the regression independently. Thus, we run the following regression:

$$y_{it}={{\alpha }_{1}y}_{it-1}+{\beta }_2{FDI}_{it}+{{\beta }_{3}EF}_{it}+{\beta }_4{CF}_{it}+{{{\beta }_5{PG}_{it}+\beta }_{6}LF}_{it}+{{\beta }_{7}LP}_{it}+{{\beta }_8{FDI}_{it}\ast LP}_{it}+{\epsilon }_{it}$$

(4)

$$y_{it}={{\alpha }_{1}y}_{it-1}+{\beta }_2{FDI}_{it}+{{\beta }_{3}EF}_{it}+{\beta }_4{CF}_{it}+{{{\beta }_5{PG}_{it}+\beta }_{6}LF}_{it}+{{\beta }_{7}LP}_{it}+{{\beta }_8{FDI}_{it}\ast EF}_{it}+{\epsilon }_{it}$$

(5)

4. Empirical Results

Before conducting the GMM tests, we examine the stationarity of the data series using the Levin-Lin-Chu (LLC) test (Levin et al. 2002) and Im-Pesaran-Shin (IPS) tests (Im et al. 2003), along with the Fisher-ADF and Fisher-PP tests developed by Maddala and Wu (1999) and Choi (2001).

Table 2. Panel data unit root test result.

			Chi-Squared Test
Variable	LLC	IPS	ADF-Fisher	PP-Fisher
GDP	−20.6888 *	−3.1327 *	764.4010 *	32.5363 *
FDI	−51.2237 *	−24.3951 *	808.5611 *	34.9455 *
Economic Freedom (EF)	−31.0783 *	−15.7595 *	940.1037 *	42.1217 *
Capital Cumulation (CF)	−10.7141 *	−17.6577 *	716.8125 *	29.9402 *
Population Growth	−0.00042	−0.0083	310.1695 *	7.7560 *
Labor Force Participation Rate (LF)	−77.5358 *	−41.1428 *	1246.3047 *	56.2808 *
Labor Productivity (LP)	−36.9462 *	−4.9398 *	720.6632 *	30.1502 *

Notes: * rejects the null hypothesis of unit root; lags in each test were defined by the Akaike information criterion.

presents the results of these tests. At the 5 percent significance level, the LLC, IPS, and both Fisher tests provide strong evidence that the six series, GDP, FDI, EF, CF, LF, and LP, are stationary, while only two statistics LLC and IPS show that population growth is of the I(1) process.

Growth regressions are outlined with a series of models: pooled OLS, fixed effects, difference GMM, and system GMM. To apply dynamic panel data models with robust properties, we utilize both two-step difference GMM and system GMM to deal with heteroskedasticity and simultaneity biases likely to exist among the variables.5 According to Nickell (1981), consistent regression coefficients resulting from difference GMM and system GMM should lie within the coefficients stemming from pooled OLS (upper bound; upward bias) and fixed effects estimators (lower bound; downward bias).

Table 3. Preliminary regression results.

	Pooled OLS	Fixed Effect	Difference GMM	System GMM
Lagged GDP (log)	0.9993 ***	0.7454 ***	0.7656 ***	0.9969 ***
	(0.0010)	(0.0367)	(0.1464)	(0.0065)
FDI as % of GDP	0.0008 **	0.0014 ***	0.0012	0.0024 **
	(0.0004)	(0.0005)	(0.0008)	(0.0011)
Economic Freedom	0.0107 ***	0.0090	0.0137	0.0347 ***
	(0.0038)	(0.0086)	(0.0114)	(0.0087)
Capital Cumulation as % of GDP	0.0010 ***	0.0015 ***	0.0014 *	0.0019 **
	(0.0003)	(0.0005)	(0.0008)	(0.0010)
Population Growth	0.0033 ***	−0.0015	0.0007	−0.0050
	(0.0012)	(0.0025)	(0.0006)	(0.0038)
Labor Force Participation Rate (log)	−0.0028	0.1889 ***	0.1421	−0.0776
	(0.0134)	(0.0725)	(0.2552)	(0.0557)
Labor Productivity (log)	−0.0091 ***	0.2607 ***	0.1718	−0.0320 ***
	(0.0019)	(0.0605)	(0.2549)	(0.0076)
AR(1) test (p-value)			0.040	0.007
AR(2) test (p-value)			0.147	0.157
Hansen Test (p-value)			0.135	0.194
Number of Observations	840	840	756	840
Number of Countries	84	84	84	84
Number of Instruments			67	74

Notes: The dependent variable is the log of GDP. Robust standard errors are in parentheses. Time dummies are included in all the regressions (not reported). *, **, and *** indicate that the coefficients are significant at the 10, 5, and 1 percent levels. Column 3 reports the results of the two-step Arellano and Bond (1991) difference GMM. Column 4 shows the results of the two-step Blundell and Bond (1998) system-GMM estimator with Windmeijer finite-sample correction. For both models, the instruments are the economic freedom index, population growth, and labor productivity. To avoid the problem of instrument proliferation and consequently overfitting of the endogenous explanatory variables, we collapse the number of instruments by dropping deeper lags as instruments (Roodman 2009).

demonstrates regression results using pooled OLS, fixed effects, difference GMM, and system GMM estimators. We use Hansen (1982) J-test to verify GMM consistency that the instruments are valid. The null hypothesis assumes that the instrumental variables satisfy the conditions, and the alternative hypothesis means that the instruments are not entirely exogenous or have been incorrectly excluded from the model (Baum et al. 2003). The Hansen test results suggest that the null hypothesis is not rejected, thus allowing the use of the GMM models. We also consider the test of second-order serial correlation of the error term suggested by Arellano and Bond (1991), in which the null hypothesis of no serial correlation of the error term is not rejected, implying the GMM estimator is valid.

The coefficients on the lagged dependent variables in all four models have a value of less than or close to one, providing strong evidence of conditional convergence. The difference GMM estimate coefficient for the lagged dependent variable is close to the fixed effects estimator, which implies that the difference GMM estimate is somewhat downward biased. The system GMM estimator with the coefficient of the lagged dependent variable falling between pooled OLS and difference GMM estimators is much closer to the upper bound of the pooled OLS estimate, provides consistent and efficient estimates, overcomes the endogeneity problem, and is a better fit for the selected panel data.

The coefficient generated by the system GMM estimator shows that economic freedom estimates are positive and significant, confirming the positive relationship between economic freedom and growth. Capital investment from FDI inflows and domestic accumulation also drive economic growth, and both sources of capital formation are equally effective in explaining economic growth. While the evidence shows that FDI leads to economic growth, a multitude of factors can still affect this relationship. Hall and Jones (1999) find that social infrastructure drives productivity and growth success through high physical and human capital investment. Differences in social infrastructure across countries can cause differences in capital accumulation, educational attainment, and productivity, which leads to differences in income between countries, thus affecting long-run economic performance concerning productivity and investment.6

The coefficients of population growth and labor force participation rate are negative but not statistically significant, while the coefficient on labor productivity is negative and statistically significant with system GMM estimates. This could be partially explained by Meyer and Sinani (2009) who find a curvilinear relationship between productivity spillovers and the level of economic development. The sample comprises developed, developing, and less developed countries, so the link between productivity and growth could thus be negative or positive.

Table 4. Regression results with the interaction terms.

	(1)	(2)
Lagged GDP (log)	0.9981 ***	0.9979 ***
	(0.0047)	(0.0058)
FDI as % of GDP	−0.0158 *	−0.0092 *
	(0.0083)	(0.0053)
Economic Freedom	0.0243 **	0.0238 *
	(0.0117)	(0.0140)
Capital Cumulation as % of GDP	0.0023 ***	0.0015 *
	(0.0008)	(0.0008)
Population Growth	−0.0002	0.0023
	(0.0029)	(0.0038)
Labor Force Participation Rate (log)	−0.0232	−0.0438
	(0.0332)	(0.0422)
Labor Productivity (log)	−0.0283 ***	0.0241 ***
	(0.0061)	(0.0075)
FDI × LP	0.0017 **
	(0.0008)
FDI × EF		0.0015 **
		(0.0007)
AR(1) Test (p-value)	0.011	0.026
AR(2) Test (p-value)	0.058	0.078
Hansen Test (p-value)	0.121	0.142
Number of Observations	840	840
Number of Countries	84	84
Number of Instruments	53	66

Notes: The dependent variable is the log of GDP. Robust standard errors are in parentheses. Time dummies are included in all the regressions (not reported). *, **, and *** indicate that the coefficients are significant at respective 10, 5, and 1 percent levels. Both columns report two-step Blundell and Bond (1998) system-GMM estimator with Windmeijer finite-sample correction. Labor force participation rate and population growth are regarded as exogenous; the other variables are treated as endogenous in the estimation and are instrumented with their lagged values. The number of instruments collapsed by dropping deeper lags as instruments to avoid the problem of instrument proliferation and overfitting of the endogenous explanatory variables (Roodman 2009).

presents the results of each regression using different indicators to assess the effect of FDI. Column (1) uses labor productivity as the indicator, while column (2) uses economic freedom. The key finding is that the coefficients of both interaction terms are statistically significant, regardless of whether FDI inflows are included as a separate regressor.7 The marginal effect of FDI on economic growth appears to depend on the levels of productivity and economic freedom. FDI inflows alone appear to be significant but negative in general. This may be attributed to the fact that, without advancement in economic freedom, FDI inflows do not contribute positively to growth.8 Without improvements in productivity, FDI inflows may not channel efficiently and could even hinder growth. The findings indicate that there is a productivity spillover effect, and that economic freedom enhances the relationship between FDI inflow and growth. However, it is important to note that our main analysis suggests that the positive impact of FDI on growth tends to increase steadily with higher levels of labor productivity and economic freedom.9 However, a certain level of labor productivity and economic freedom may be necessary before the benefit of FDI can be fully realized in host countries. Only those countries that possess adequate absorptive capacity and implement progressive and transparent economic policies will likely reap the advantages of FDI inflows.

Table 5. (a). Regression results with alternative specifications: non-OECD countries. (b). Regression results with alternative specifications: OECD countries.

(a)
	(1)	(2)	(3)
Lagged GDP (log)	0.9978 ***	0.9945 ***	0.9987 ***
	(0.0059)	(0.0054)	(0.0032)
FDI as % of GDP	0.0014	−0.0150 *	−0.0106 *
	(0.0009)	(0.0084)	(0.0064)
Economic Freedom	0.0220 ***	0.0246 ***	0.0128
	(0.0075)	(0.0121)	(0.1058)
Capital Cumulation as % of GDP	0.0015 *	0.0020 ***	0.0032 ***
	(0.0009)	(0.0010)	(0.0012)
Population Growth (%)	−0.0004	−0.0003	−0.0003
	(0.0041)	(0.0033)	(0.0026)
Labor Force Participation Rate (log)	−0.0602	−0.0332	0.0078
	(0.0419)	(0.0342)	(0.0416)
Labor Productivity (log)	−0.0290 ***	−0.0332 ***	−0.0184 *
	(0.0133)	(0.0108)	(0.0106)
FDI × LP		0.0016 ***
		(0.0008)
FDI × EF			0.0068 **
			(0.0034)
AR(1) Test (p-value)	0.017	0.040	0.044
AR(2) Test (p-value)	0.150	0.094	0.089
Hansen Test (p-value)	0.237	0.355	0.173
Number of Observations	500	500	500
Number of Countries	50	50	50
Number of Instruments	47	49	38
(b)
	(1)	(2)	(3)
Lagged GDP (log)	0.9883 ***	0.9886 ***	0.9931 ***
	(0.0066)	(0.0079)	(0.0055)
FDI as % of GDP	0.0028 **	−0.0142 **	−0.0153 **
	(0.0011)	(0.0067)	(0.0060)
Economic Freedom	0.0441	0.0113	−0.0061
	(0.0457)	(0.0194)	(0.0247)
Capital Cumulation as % of GDP	0.0016 *	0.0012 *	0.0014 *
	(0.0009)	(0.0007)	(0.0007)
Population Growth	0.0119 *	0.0118 *	0.0071 *
	(0.0066)	(0.0061)	(0.0042)
Labor Force Participation Rate (log)	−0.1143	−0.0577	−0.0061
	(0.0942)	(0.0546)	(0.0396)
Labor Productivity (log)	−0.0076	−0.0064	−0.0082
	(0.0125)	(0.0091)	(0.0098)
FDI × LP		0.0014 **
		(0.0006)
FDI × EF			0.0022 ***
			(0.0008)
AR(1) Test (p-value)	0.033	0.032	0.064
AR(2) Test (p-value)	0.149	0.173	0.103
Hansen Test (p-value)	0.291	0.239	0.129
Number of Observations	340	340	340
Number of Countries	34	34	34
Number of Instruments	34	27	29

Notes: The dependent variable is the log of GDP. Robust standard errors are in parentheses. Time dummies are included in all the regressions (not reported). *, **, and *** indicate that the coefficients are significant at respective 10, 5, and 1 percent levels. All three columns report two-step Blundell and Bond (1998) system-GMM estimator with Windmeijer finite-sample correction. Labor productivity, labor force participation rate, and population growth are predetermined variables; foreign direct investment and capital accumulation are treated as endogenous in the estimation and are instrumented with their lagged values. The number of instruments collapsed by dropping deeper lags as instruments to avoid the problem of instrument proliferation and overfitting of the endogenous explanatory variables (Roodman 2009).

a,b presents the results of models estimated using the system GMM, separated into subsets for OECD and non-OECD countries. In the regressions that exclude interaction terms, the coefficients for FDI are positive and statistically significant. Notably, the FDI coefficient is approximately twice as large for OECD countries compared to non-OECD countries. Additionally, the coefficients for domestic capital formation are significant and roughly equal for both groups.

The interaction terms indicate a strong positive relationship between FDI inflows and productivity, as well as between FDI inflows and economic freedom, for both transition economies and OECD countries. Although the coefficient estimates for the FDI-productivity interaction term are similar across both groups, the interaction between FDI inflows and economic freedom is more pronounced in OECD countries compared to non-OECD countries. This discrepancy may suggest that the interaction term captures a general trend where FDI contributes more significantly to the growth of transition and OECD countries that possess a minimum threshold of economic infrastructure. For OECD countries, greater economic freedom appears to enhance their absorptive capacity, which is crucial for economic growth driven by FDI inflows.

Economic freedom primarily drives economic growth in non-OECD countries. However, when interaction terms are included in the regression analysis, FDI inflows can have a negative effect on growth. Generally, FDI inflows are beneficial due to productivity spillover effects, which appear to be similar for both OECD and transition economies. OECD countries tend to have a greater capacity to absorb these FDI spillovers, particularly when economic freedom is improved. Nevertheless, economic freedom alone does not ensure robust economic growth in OECD countries; its impact on enhancing economic growth is more significant in non-OECD countries.

The average population growth rate in non-OECD countries is three times higher than in OECD countries. While a larger population can stimulate economic growth in OECD countries, it may place a significant economic burden on non-OECD countries. The effect of labor force participation on growth is negative but statistically insignificant for both OECD and non-OECD countries. Furthermore, the results regarding labor productivity are mixed for both groups, although it seems to have a more detrimental impact on economic growth in non-OECD countries.

Figure 2 presents data from 2010 to 2019 to investigate the relationship between economic growth and FDI inflows in OECD and transition economies. The fitted lines reveal a significantly stronger positive correlation between FDI and economic growth in OECD countries than in non-OECD countries. While many non-OECD countries are currently in transition and experiencing relatively high economic growth, OECD countries appear to gain more advantages from FDI inflows. Figure 3 illustrates the relationship between economic freedom and growth, distinguishing between OECD and transition economies. The fitted lines show a weak positive correlation between output growth and economic freedom in non-OECD countries, while a negative correlation is observed in OECD countries. Although OECD nations typically create a more favorable environment for economic freedom, their output growth appears to decrease in response to changes in this freedom.

5. Conclusions

There is a consensus among policymakers globally that FDI inflows are a vital source of external financing for developing countries, emerging economies, and nations undergoing transition (UNCTAD 2019). FDI inflows are instrumental in transferring capital and technology, enhancing productivity, and driving economic growth. They supply essential capital for economic development and play a significant role in job creation, particularly in economies advancing toward higher levels of development. In countries transitioning between stages of development, economic openness fosters an environment that facilitates effective capital allocation, ultimately resulting in increased productivity and economic expansion. Furthermore, productivity improvements associated with FDI inflows can often be attributed to the introduction of new capital and technology. Several factors may explain the relationship between FDI inflows and enhanced productivity levels.

We utilized the GMM system estimator to analyze a panel of 84 countries spanning the years 2010 to 2019. Our study investigates the productivity spillover effects of FDI inflows, as well as the role of economic freedom in attracting foreign investment. By categorizing the countries into two groups—transition economies and OECD economies—we focused on the impact of FDI inflows on economic growth and the mechanisms through which this influence operates, especially in countries at varying stages of development. Our empirical findings reveal that FDI inflows generally provide significant productivity spillover benefits, which are relatively comparable between OECD and transition economies. Economic freedom is a critical factor in attracting FDI inflows; OECD countries with higher levels of economic freedom tend to exhibit greater absorptive capacity for FDI spillovers. Nonetheless, economic freedom alone does not guarantee robust economic growth in OECD nations; its effects are more pronounced in fostering economic growth in non-OECD countries. These results further corroborate earlier studies, affirming that only economies with adequate levels of economic infrastructure and development can fully capitalize on the benefits of FDI inflows.

This paper investigates the general spillover effects of FDI and enriches the existing literature by examining the relationship between FDI presence, productivity, and economic freedom. The study focuses explicitly on distinct groups of transition economies and OECD countries. Although there are policies aimed at attracting foreign investment, empirical evidence regarding the growth effects of FDI inflows remains influenced by the unique conditions of each country. Furthermore, when different definitions of spillover variables are employed, the results can vary significantly across nations. This underscores an important area for further research to explore how levels of economic freedom and stages of development in countries shape this overall relationship and its evolution over time.