Foreign Direct Investment Inflows and Green Technology Innovation: The Role of Institutional Quality

The world is facing a dual challenge of our time: sustaining economic growth while confronting the increasingly severe consequences of climate change and environmental degradation. Landmark international commitments such as the Paris Agreement on climate change and the United Nations' Sustainable Development Goals (SDGs) have created a global impetus, pushing nations to transition towards a green and sustainable economic model [1, 2]. In this transition, green technology innovation (GTI) is considered a foundational solution, expected to break the historical link between economic growth and resource depletion, allowing economies to develop without environmental costs [3, 4].

For emerging economies, particularly those in the Association of Southeast Asian Nations (ASEAN), this challenge is even more pressing. The ASEAN-5 region (comprising Indonesia, Malaysia, Philippines, Thailand, and Vietnam) has been witnessing impressive economic growth, largely driven by abundant foreign direct investment (FDI) inflows [5]. However, this growth trajectory has also been accompanied by severe environmental pressures, from air and water pollution to the depletion of natural resources [6, 7]. This raises a core policy question: how can these nations leverage external resources, especially FDI, to "green" their development paths and achieve sustainable prosperity?

Theoretically, FDI inflows are seen as a vital conduit, bringing not only capital but also advanced technology and managerial knowledge to host countries. In the environmental sphere, the academic debate on the impact of FDI on GTI is divided into two main schools of thought. On one hand, the "Pollution Halo Hypothesis" posits that FDI from developed countries often brings cleaner, more energy-efficient production technologies and higher environmental management standards than those of domestic firms [8-10]. The presence of these multinational corporations (MNCs) can create green technology spillover effects through competition, supply chain linkages, and the mobility of skilled labor, thereby boosting the green innovation capacity of the entire economy [11].

On the other hand, the "Pollution Haven Hypothesis" presents a contrasting scenario. This hypothesis argues that, to evade stringent environmental regulations in their home countries, MNCs tend to relocate resource-intensive and polluting industries to countries with laxer environmental regulations [12, 13]. In this case, FDI not only fails to promote but may even hinder the process of GTI and exacerbate environmental problems in the host country. The coexistence of these two hypotheses, along with conflicting empirical results in previous studies, suggests that the relationship between FDI and GTI is a complex and context-dependent interaction.

This inconsistency suggests that the impact of FDI is deeply dependent on the specific conditions of the host country, with institutional quality (IQ) being considered a key moderating factor. Based on institutional theory, the formal (e.g., laws, regulations) and informal (e.g., social norms) "rules of the game" strongly shape the incentives and behaviors of economic agents, including MNCs [14]. A sound institutional environment—characterized by the rule of law, strong protection of intellectual property rights, low levels of corruption, and government effectiveness—reduces transaction costs and risks for economic activities, thereby encouraging MNCs to make long-term investments, transfer advanced technology, and engage in R&D activities in the host country [15-17]. Conversely, weak institutions create an unstable environment, encouraging short-term, resource-extractive investments and rent-seeking behaviors that are detrimental to the environment.

However, existing studies examining the role of institutions have predominantly viewed it as an amplifying factor—a mechanism that enhances an assumed pre-existing positive impact of FDI [18]. This study challenges that traditional "amplification" view by asking a more fundamental question: what if IQ is not merely a factor that adjusts the intensity of FDI's impact, but a prerequisite that determines whether a positive impact can emerge in the first place? Instead of just asking how much institutions affect green technology transfer, we investigate whether the expected green benefits from FDI can be realized in the absence of a sufficiently strong institutional framework.

Based on the identified research gap and the theoretical foundation analyzed, this paper sets two specific research objectives:

(i) To re-examine the direct impact of FDI inflows on GTI in the ASEAN-5 countries.

(ii) To analyze the moderating role of IQ in the relationship between FDI and GTI, specifically testing the hypothesis that high-quality institutions are a prerequisite for activating the positive impact of FDI.

By addressing these objectives, this study expects to contribute new empirical evidence that clarifies the complex role of IQ within the specific context of the ASEAN-5 economies. These findings will not only have academic significance but also provide important implications for ASEAN policymakers in their efforts to forge a sustainable development path.

To mitigate the influence of outliers that could bias the regression results, all continuous variables in the model were winsorized at the 1st and 99th percentiles. Winsorizing at the 1% and 99% levels affects a maximum of 2 observations at each tail of the distribution for each variable. Specifically, 4 observations (2 at the lower tail and 2 at the upper tail) of the FDI variable and 3 observations of the Trade_Openness variable had their values adjusted. The remaining variables did not have extreme values exceeding the winsorization threshold.

GTI (GTI_log): Following the convention in innovation studies [33], we use patent data as a proxy for the output of innovation activities. Specifically, GTI is measured by the number of patents in environment-related technology fields filed by inventors from the host country, according to the OECD classification. We acknowledge the limitations of this measure: not all innovations are patented, and the economic value of patents varies. However, it is an objective, quantifiable, and comparable measure across countries, reflecting formal inventive efforts in the green technology sector. Taking the logarithm of (number of patents + 1) is a standard technique to handle the issue of zero values and reduce the skewness of the data distribution.

IQ: IQ is a multidimensional concept. Instead of choosing a single indicator that could lead to omitted variable bias, we apply Principal Component Analysis (PCA) to construct a composite index. This method extracts common information from the six component indicators of the WGI dataset [34]. The first principal component (PC1) explains 71.3% of the total variance with an eigenvalue of 4.28, indicating that the six WGI indicators are highly correlated and that PC1 effectively captures the common dimension of IQ. The factor loadings of PC1 range from 0.35 to 0.45, specifically: Government Effectiveness has the highest value (0.45), followed by Rule of Law (0.44), Regulatory Quality (0.43), Voice and Accountability (0.42), Control of Corruption (0.41), and Political Stability (0.35). The relative uniformity of the loadings suggests that all six institutional dimensions contribute significantly to the composite index.

FDI: The FDI variable is measured as net FDI inflows as a percentage of GDP. This is a standard international measure that reflects the relative importance of FDI to the size of the host economy, allowing for a meaningful comparison of the economic significance of these capital flows across countries with different GDP sizes.

The table below summarizes the variables used in the study.

3.3 Analysis procedure and estimation strategy

The analysis procedure is designed to systematically address econometric challenges, ensuring that the estimation results are robust and reliable.

Preliminary Analysis: Before proceeding with regression, we conduct descriptive statistical analysis to grasp the basic characteristics of the data (mean, standard deviation, min, max) and correlation matrix analysis to preliminarily examine the relationships between variables and detect potential signs of multicollinearity.

Static Panel Estimations and the Issue of Cross-Sectional Dependence: We begin with traditional static panel models, including Pooled OLS, Fixed Effects (FE), and Random Effects (RE) models. The Hausman test is used to choose between FE and RE. However, a potential concern in panel data analyses, especially with country-level samples, is the presence of cross-sectional dependence. This occurs when macroeconomic shocks or specific events simultaneously affect all countries in the sample. If not controlled for, this phenomenon can lead to biased standard error estimates. Therefore, we will use the Pesaran (2004) test to detect cross-sectional dependence. If this phenomenon exists, to thoroughly address the problems of heteroskedasticity, autocorrelation, and cross-sectional dependence, we will re-estimate the FE model using Driscoll–Kraay [35] standard errors. This method produces standard errors that are robust and consistent even in the presence of very general forms of cross-sectional dependence.

Main Estimation using System GMM developed by Arellano and Bover [36] and Blundell and Bond [37]: We recognize that applying System GMM with N = 5 countries requires special caution. However, the choice of this method remains appropriate for three reasons. First, System GMM is designed to address endogeneity in dynamic panel data, a core feature of the technological innovation process that static methods cannot handle. Second, with T = 20 years, the relatively high T/N ratio helps mitigate the standard error issues commonly found in small-N samples. Third, many macroeconomic studies have successfully applied GMM to small country samples when there is a strong theoretical basis for dynamics and endogeneity [38].

To ensure the robustness of the GMM estimates, we perform two crucial specification tests: (i) The Sargan/Hansen test: This tests the null hypothesis H₀ that the instruments are valid (uncorrelated with the error term); (ii) The Arellano-Bond test for autocorrelation: This checks for the presence of autocorrelation in the differenced errors. The model is considered well-specified if there is evidence of first-order autocorrelation (AR(1)) but no evidence of second-order autocorrelation (AR(2)). We also implement the following control measures: (i) limiting the depth of instruments to a maximum of 2 lags to avoid the problem of too many instruments; (ii) using the collapse option to reduce the instrument matrix; and (iii) including time dummies in the model to control for common shocks. The FDI variable and the FDI × IQ interaction term are treated as endogenous variables, with their lags from t-2 and earlier used as instruments in the difference equation. Additionally, we supplement with two reference estimators. The first is the bias-corrected LSDV (LSDVC) model by Bruno [39], suitable for dynamic panel data with small N. The second is a dynamic FE model with Driscoll-Kraay standard errors.

Robustness Checks: To answer the question of whether the results are sensitive to choices of model and variables, we will perform a series of robustness checks: (i) Using an alternative measure: To check if the results depend on how the key variable is measured, we will re-estimate the model using an alternative measure for IQ, such as using only the "Rule of Law" index from WGI; (ii) Changing the set of control variables: We will sequentially remove each control variable from the model to check if the coefficients of the main variables (FDI, IQ, and the interaction term) change significantly; (iii) Checking for non-linear relationships: To explore the possibility of more complex relationships, we will test for a non-linear threshold by adding the square of the FDI variable (FDI²) to the model.

All estimations and tests in this study are performed using the specialized statistical software Stata version 17.

5.1 Discussion of the main results

The first core finding of this study is the absence of a statistically significant direct impact of FDI inflows on GTI in the ASEAN-5 countries during the 2002-2021 period. The results from the System GMM model (Table 5) show that the coefficient of the FDI variable is not only statistically insignificant but also has a slight negative sign. This finding directly challenges the optimistic assumption of the "Pollution Halo Hypothesis," which posits that FDI is inherently a conduit for clean technology and advanced environmental standards [8, 9]. Our results suggest that the mere act of attracting FDI is not a sufficient condition to generate the expected positive technological impacts. These benefits do not arise automatically but require catalytic factors from the host country's institutional environment.

Conversely, this result provides a more fitting perspective with the 'Pollution Haven Hypothesis,' but with a more complex nuance. It does not necessarily assert that all FDI into ASEAN-5 is polluting, but suggests that in an average institutional environment, FDI flows tend not to prioritize green innovation objectives. This aligns with previous studies that found negative or insignificant impacts of FDI on the environment in developing countries with lax regulations [13, 27].

The difference between our results and some studies that found a direct positive impact [18] can be explained by the specific context of the ASEAN-5. Unlike developed economies (OECD) where similar studies are often conducted, the ASEAN-5 countries during the study period were still in the process of refining their legal frameworks for the environment and intellectual property. In many cases, the main drivers for attracting FDI to this region were low labor costs, large market size, and favorable access to resources, rather than an innovative environment. Consequently, MNCs may lack sufficient incentive to transfer their most advanced green technologies, which require high investment costs and a sufficiently secure legal environment to protect intellectual property. Instead, they might only apply technology that just meets the minimum environmental standards of the host country, leading to negligible green technology spillover effects.

5.2 Discussion of the moderating role of IQ and the significance of the threshold

The most important finding and also the most outstanding contribution of this study is the robust demonstration that IQ acts as a determinant moderator. The positive and highly statistically significant coefficient of the FDI × IQ interaction term (0.150, p < 0.01) shifts the research question from "Does FDI promote GTI?" to "Under what institutional conditions does FDI promote GTI?". This result extends North's [14] institutional theory into the field of environmental economics, affirming that the "rules of the game" not only shape general economic behavior but also determine the quality and technological orientation of foreign investments.

The mechanism behind this moderating role can be explained as follows:

(i) Reducing risk and transaction costs: A good institutional environment, characterized by the Rule of Law, effective control of corruption, and strict protection of intellectual property rights, significantly reduces the risks for MNCs when transferring core, expensive, and sensitive technologies. They will be more confident that their intellectual assets will not be stolen and that contracts will be enforced, thereby encouraging long-term investments in R&D and green technology.

(ii) Creating a level playing field and compliance pressure: High-quality institutions are often accompanied by clear environmental regulations and effective enforcement mechanisms. This eliminates the "competitive advantage" of polluting firms, forcing all companies, including MNCs, to compete based on efficiency and innovation. This compliance pressure encourages MNCs to adopt the cleaner technologies they already use in their home countries, creating a positive spillover effect.

The core point of the study is the identification of a threshold value for IQ (IQ ≈ 0.1). This is not just a statistical figure; it carries profound economic meaning. Since the WGI index is standardized with a global mean of 0, the threshold of 0.1 implies that an ASEAN-5 country needs to build a sufficiently strong institutional foundation that surpasses the world average. Only then can the country effectively absorb and leverage the green technology spillover effects from FDI.

For instance, in countries with IQ < 0.1 (e.g., an IQ index of -0.5), the marginal effect of FDI on GTI would be: -0.015 + 0.150 × (-0.5) = -0.09. In this case, an MNC might decide to build a factory in this country to take advantage of cheap labor. However, due to concerns about copyright infringement and lax environmental regulations, they only transfer an older, more energy-intensive, and polluting production line. FDI flows in, GDP increases, but the nation's green innovation capacity not only fails to improve but may even be stifled.

Or, for countries with IQ > 0.1 (e.g., an IQ index of 0.8, equivalent to countries like Malaysia in recent years), the marginal effect of FDI on GTI would be: -0.015 + 0.150 × (0.8) = +0.105. In this country, MNCs perceive that laws are strictly enforced and intellectual property rights are well-protected. To meet rising environmental standards and compete in the market, they decide to invest in an R&D center and apply the most advanced green production technology. Local engineers and suppliers gain access to and learn from this technology, creating a spillover effect that boosts the entire nation's green innovation ecosystem.

The fact that the average IQ of the research sample (-0.211) lies below the 0.1 threshold explains why the average direct impact of FDI was not statistically significant. It implies that, for most of the study period, the "typical" ASEAN-5 country had not yet met the necessary institutional conditions to turn FDI into a driver for green innovation. The robustness check result in column (6), using the "Rule of Law" (RL) index instead of the composite IQ, further strengthens this argument. It shows that legal assurance is one of the most core aspects of institutions for attracting high-quality FDI.

5.3 Discussion of the non-linear relationship test results

The study also tested for the possibility of a non-linear (U-shaped or inverted U-shaped) relationship between FDI and GTI by adding the FDI² term to the model, but the results showed that the coefficient of this term was not statistically significant (Table 6, column 9). The failure to find this non-linear relationship in the ASEAN-5 context can be explained by several reasons.

First, it is possible that the ASEAN-5 economies have not yet reached a sufficiently large scale of FDI for non-linear effects to become apparent. Hypotheses about non-linear relationships often suggest that after a certain FDI threshold, benefits may diminish (due to excessive competition weakening domestic firms) or accelerate (due to reaching a critical mass to create innovative industrial clusters). It may be that FDI inflows into the region, though large in absolute terms, are still not sufficient in terms of GDP share and quality to trigger these complex dynamics.

Second, and more importantly, this result further reinforces the main argument of the study: the decisive factor is not the "quantity" of FDI, but the "interaction" between FDI and IQ. The linear interaction model has captured the main dynamic governing this relationship very well. This means that, instead of the impact of FDI changing with its own scale, the impact of FDI changes consistently with the improvement of IQ. The fundamental relationship is a function of IQ, not a function of FDI scale. Therefore, adding the FDI² term did not bring significant additional explanatory power to the model. This suggests that for ASEAN policymakers, the focus should not be on "attracting more FDI at all costs" but on "improving institutions to enhance the quality of existing and future FDI flows".

5.4 Limitations and future research directions

Despite achieving its proposed research objectives, we acknowledge that our study still has some limitations. Specifically, the use of aggregate data at the national level for both FDI and patent counts may obscure important differences in the nature of capital flows (e.g., FDI by industry) and the true value of each innovation.

Future research could delve deeper by disaggregating FDI by sector to identify which types of investment truly deliver green benefits. Furthermore, shifting to an industry- or firm-level analysis would allow for a direct test of technology spillover mechanisms and the exploration of more diverse measures of green innovation beyond patent data. These directions would provide a more detailed picture of how to build an effective institutional framework to optimize the benefits from FDI.