Abstract

Translate

Amid escalating concerns over air pollution and demographic shifts, industrial robots have emerged as a key solution to enhancing energy efficiency, reducing emissions, and fostering economic growth. However, existing research often overlooks their role in shaping green total factor productivity (GTFP), a critical measure of environmentally sustainable economic performance. This study investigates the relationship between industrial robot applications (IRAs) and high-quality economic development (HQED) by integrating theoretical modeling and empirical analysis. Using panel data from 32 countries (16 developed and 16 developing) over the period of 1993–2019, classified according to the 2023 International Monetary Fund (IMF) standards, this study employs fixed-effects models, system generalized method of moments (SYS-GMM), and threshold regression models to assess IRA-induced impacts on HQED. The findings reveal that IRAs significantly contribute to HQED, with a stronger effect observed in developing economies. Moreover, a threshold effect exists, wherein environmental regulations (ERs) mediate the effectiveness of IRAs in improving GTFP. Additionally, IRAs drive HQED through foreign direct investment (FDI) and technological innovation (TI). These results provide empirical evidence and policy insights for leveraging industrial automation to promote sustainable economic growth across different national contexts.

Full text

Turn on search term navigation

Translate

1. Introduction

In recent decades, the global economy has faced a dual imperative: sustaining economic growth while addressing increasing ecological pressures. Traditional growth models, heavily dependent on capital and labor inputs, have yielded diminishing returns and contributed to environmental degradation. This context has intensified the demand for new development paradigms that harmonize efficiency and sustainability. Within this framework, green total factor productivity (GTFP) has emerged as a crucial metric for evaluating both economic efficiency and environmental performance. Based on the classical total factor productivity (TFP) theory proposed by [1], GTFP integrates undesirable outputs—such as pollution and energy inefficiency—into productivity assessment, aligning long-term growth objectives with ecological constraints.

Industrial robots (IRAs) have increasingly become an essential component of high-quality economic development (HQED). As a representative form of intelligent automation, IRAs contribute to improving operational productivity, optimizing resource utilization, and reducing environmental impacts [2,3]. By enabling labor substitution, lowering energy consumption, and enhancing precision in production processes, IRAs foster cleaner production and more sustainable industrial practices. Beyond their technical contributions, IRAs also enhance workplace safety and reduce the dependence on volatile labor supply chains, thereby strengthening economic resilience.

Although numerous studies have explored the effects of technological progress on economic growth [4,5], existing research tends to understate the significance of IRAs in fostering GTFP, especially in the context of global efforts toward sustainability. Furthermore, prior analyses often overlook the differentiated impact of IRAs across countries with varying development levels. Developed and developing economies exhibit distinct industrial structures, environmental regulation frameworks, and technological innovation capacities, which influence the scale, scope, and effectiveness of IRA implementation.

From the perspective of environmental regulation, cross-country differences in institutional design and enforcement intensity lead to heterogeneous effects of IRAs on green productivity. Environmental regulations (ERs), while traditionally perceived as constraints on productivity, can also incentivize technological innovation and cleaner production practices [6]. Consequently, the relationship between IRAs and GTFP may exhibit threshold characteristics depending on the regulatory environment, as suggested by previous empirical studies [7,8,9]. The role of ERs as a moderating factor thus warrants systematic investigation.

Moreover, the mechanisms through which IRAs affect HQED involve several interrelated factors. Foreign direct investment (FDI) serves as a channel through which advanced automation technologies and energy-efficient practices are transferred across borders, enhancing local innovation and industrial upgrading [7,8]. Additionally, research and development (R&D) plays a crucial role in enabling green technological advancement and promoting GTFP, especially in emissions-intensive sectors [9,10,11]. Technological innovation (TI) further mediates this relationship by facilitating structural transformation and knowledge-intensive production, as documented by [12,13,14].

This study investigates the influence of IRAs on high-quality economic development from a global perspective. It contributes to the literature in three ways. First, it refines conventional economic growth models by integrating scale effects and the pricing dynamics of automation technologies. Second, it conducts a comprehensive empirical analysis using panel data from 32 countries—16 developed and 16 developing economies—spanning the period of 1993–2019, following the classification standards of the International Monetary Fund [11]. Third, it explores the mediating roles of ERs, FDI, TI, and R&D in shaping the impact of IRAs on GTFP, with a focus on heterogeneity across development levels.

By incorporating nonlinear modeling approaches, including threshold regression and system generalized method of moments (SYS-GMM), this study uncovers the differentiated effects of IRAs under varying environmental and institutional conditions. It provides empirical evidence that supports the hypothesis that IRAs significantly contribute to HQED, particularly through their interactions with environmental regulation, foreign investment, and innovation ecosystems.

The remainder of this paper is structured as follows: Section 2 elaborates on the theoretical framework and hypotheses; Section 3 outlines the methodology and data sources; Section 4 presents the empirical findings and robustness tests; and Section 5 concludes with policy implications and academic contributions.

2. Conventional Facts, Theoretical Framework, and Hypotheses

2.1. Total Factor Productivity and Its Green Transformation

In the context of global sustainable development, GTFP has emerged as a prominent index commanding considerable attention within both academic and practical circles. GTFP research is centered on the pursuit of efficient resource utilization and the promotion of environmental sustainability alongside economic growth. It represents a metric that augments conventional TFP assessment by integrating environmental benefits [12]. This comprehensive evaluation takes into account various factors, including output, labor, capital, and environmental inputs, thereby shedding light on the environmental gains achieved throughout the production process [13]. The introduction of GTFP holds the potential to guide enterprises toward sustainable production practices and green development objectives. By striking a balance between economic growth and environmental conservation, GTFP contributes to the realization of high-quality economic development within a nation [14,15].

2.2. Classification of Countries Based on Economic Development

To analyze the impact of industrial robot applications (IRAs) on high-quality economic development (HQED), this study categorizes countries into developing and developed economies, as IRA adoption, economic structure, and technological innovation differ significantly between these two groups. Following the International Monetary Fund (IMF) classification and GDP-based standards [16], countries are divided into developed economies, characterized by advanced industrial systems, technological capabilities, and stable economic frameworks, and developing economies, which exhibit emerging industries, lower per-capita GDP, and evolving market structures. To ensure analytical consistency, this study adopts 1993–2019 as the sample period and selects 16 developed and 16 developing countries (Table 1). This selection aligns with established research methodologies [17,18] and allows for robust cross-country comparisons of HQED indicators such as green total factor productivity (GTFP), industrial structure, and innovation efficiency. The classification framework provides a structured basis for assessing how IRA adoption influences economic growth, environmental sustainability, and technological progress across varying economic conditions.

2.3. Heterogeneous Impact of IRAs on HQED in Developing and Developed Countries

This study aims to explore the impact of industrial robot applications (IRAs) on the high-quality economic development (HQED) of national economies in varying stages of development. To accurately reflect international differences, we classify countries into developed and developing groups based on the GDP-based classification standards provided by the International Monetary Fund [11] and the World Bank (2022), which are widely recognized for their methodological rigor. This classification replaces prior, simplified categorizations and ensures consistency with international development indicators.

To measure industrial automation, we calculate industrial robot density, defined as the number of industrial robots per 10,000 workers, following [19]. For assessing green economic performance, we adopt the Global Malmquist–Luenberger (GML) index based on the slacks-based measure (SBM) directional distance function (DDF), as proposed by [20], to estimate green total factor productivity (GTFP). This metric effectively captures both productivity gains and environmental constraints, serving as a comprehensive proxy for HQED.

To illustrate the relationship between IRAs and HQED, we plotted the correlation between industrial robot density and GTFP for the selected countries over the period from 1993 to 2019 (Figure 1). The fitted trend line indicates a consistent positive association across both developed and developing nations, suggesting that the widespread adoption of IRAs contributes meaningfully to green productivity improvements and sustainable economic upgrading.

2.4. Theoretical Examination and Hypothesis Development for IRAs in the Context of HQED

Productivity is the fundamental driver of national and regional economic growth, and its sustained enhancement is essential for achieving long-term stability. The transition from traditional economic growth to high-quality economic development (HQED) is inherently dependent on the continuous expansion of green total factor productivity (GTFP) [21]. In this context, understanding the impact of industrial robot applications (IRAs) on HQED requires an examination of the mechanisms through which technological progress, particularly automation and robotics, influences productivity growth across different economic conditions.

Building on endogenous growth theory, this study explores how industrial robots contribute to capital deepening and TFP augmentation, enabling economies to overcome diminishing marginal returns on capital and achieve sustained long-term growth. As intelligent technologies, IRAs enhance investment efficiency, stimulate higher savings and investment rates, and accelerate technological progress. To empirically assess these effects, we analyze panel data on 38 countries and 17 manufacturing sectors, identifying the contribution of IRAs to productivity enhancement through two key channels: capital deepening and TFP growth.

Additionally, this study investigates the threshold effects of environmental regulations (ERs) in shaping the relationship between IRAs and GTFP. Given that the economic impact of IRAs varies across national contexts, we further examine the role of technological innovation (TI), research and development (R&D), and foreign direct investment (FDI) in mediating the influence of IRAs on HQED. This approach allows for a comprehensive comparison between developed and developing economies, offering insights into how technological adoption, regulatory environments, and investment dynamics interact to shape long-term economic outcomes.

According to the model by [22], many tasks (industries) compose the economy. All tasks x are a part of the continuous system defined by the interval left–left $[N - 1, I]$ , –right, in which interval length is unaffected by the introduction of new jobs into the labor market. The total output is calculated by summing the outputs of all tasks or by assuming the Cobb–Douglas production function for each activity. Formula (1) derives the overall output:

(1) $l n Y = \int_{N - 1}^{N} l n y (x) d x,$

where

Y

represents the total output,

y (x)

signifies the output of individual tasks, and

[N - 1, I]

denotes the task interval. Each business completing task x must decide to employ industrial robots or human labor for manufacturing. For simplicity, tasks x generated by industrial robots are situated within interval

[N - 1, I]

, while tasks x produced by labor are positioned in interval

[I, N]

. The following equation (Formula (2)) can be used to elucidate the production function of each task:

(2) $y (x) = \{\begin{matrix} γ_{m} (x) k {(x)}^{a} m {(x)}^{1 - α} i f x \in [N - 1, I] \\ γ_{l} (x) k {(x)}^{n} l {(x)}^{1 - α} i f x \in [I, N] \end{matrix},$

where

γ_{m}

and

γ_{l}

stand for the productivity of production employing robots and the labor force, respectively, and

k

l

, and

m

denote capital stock, labor force, and industrial use of robots, respectively. Mainly,

γ_{m}

γ_{l}

. This study deduces from the equation of

l n Y

the general production function, as in Formula (3):

(3) $\begin{matrix} Y = A K^{α} {(\frac{M}{I - N + 1})}^{(1 - α) (I - N + 1)} {(\frac{L}{N - I})}^{(1 - α) (N - I)}, w h e r e A = \\ e x p (\int_{N - 1}^{I} (l n γ_{m} (x)) d x + \int_{I}^{N} (l n γ_{l} (x)) d x) . \end{matrix}$

Formula (3) demonstrates that a greater utilization m of industrial robots corresponds to a larger total output. The nature of this connection demonstrates nonlinearity. In practical contexts, developed nations maintain an average of 3.26 industrial robots per 10,000 workers with developing nations at an average of 1.78. This dissimilarity underscores diversity in adopting industrial robots between developed and developing countries. In summary, integrating industrial robots into economic operation bolsters economic growth. Nevertheless, the impact on economic growth displays a significant disparity due to varying degrees of incorporation of industrial robots between developed and developing nations. Currently, nations strive for economic development while considering environmental sustainability and low-carbon practices; GTFP serves as a pivotal metric for assessing the equilibrium between economic and green development, which constitutes the foundation of a high-quality economy. Thus, we posit that

H1.

IRAs positively contribute to GTFP.

Second, ERs foster the positive effects of IRAs on HQED. A longstanding belief is that effective ER management will increase production costs and render difficult the growth of production scales, which will lessen productivity. However, a revisionist school of thought contends that spending money on TI and pollution control helps not only meet consumer demand for GTFP in markets with increased environmental awareness but also reduce wasteful energy and raw material consumption [6]. Regarding the influence of ER on GTFP, previous studies propose three major hypotheses. First, ERs impede GTFP advancement. Stringent ERs will increase the expense of pollution control for local businesses and impede R&D and innovation efforts. Consequently, GTFP decreases [23,24,25]. Researchers also consider the pollution paradise hypothesis, which posits that businesses in industries with high pollution levels are frequently located in areas with lax ERs [26]. They find that polluting businesses relocate to regions with improved local environmental rules to escape high production costs, which reduces local GTFP [27]. These differences, which exert impacts on the ER efficacy, influence the policy goals of local governments [28]. Thus, we present the following:

H2.

The influence of IRAs on GTFP has the threshold effect of ERs.

Third, IRAs contribute to HQED by facilitating foreign direct investment (FDI), which plays a critical role in economic development and technological progress. Green total factor productivity (GTFP) serves as a comprehensive measure of industrial resource efficiency, and FDI has been widely recognized as a key driver of GTFP enhancement through technology spillovers [7]. By transferring energy-efficient and environmentally friendly technologies from foreign enterprises and attracting high-caliber talent, FDI improves energy utilization efficiency in host countries, thereby fostering GTFP growth [8]. Empirical studies provide strong evidence for this relationship. [16], in their study on Anglophone and Francophone African nations from 1997 to 2017, reveal that FDI positively influences economic growth, with this effect being further reinforced by the presence of IFRS. Similarly, [29] analyze panel data from 7855 enterprises and demonstrate that FDI from Hong Kong, Macao, and Taiwan significantly enhances forestry TFP in China. [30], using provincial-level data from China (2004–2020), show that FDI activities contribute to GTFP growth, with FDI stock exerting a pronounced impact on green development. Notably, their findings suggest that FDI plays an even greater role in regions with lower levels of green growth and marketization. Beyond direct technological spillovers, institutional factors also mediate the impact of FDI on GTFP. [31] examine the role of fiscal decentralization and find that greater decentralization stimulates GTFP growth. Similarly, [32] analyze outward FDI (OFDI) from China to Belt and Road Initiative (BRI) countries and highlight that the positive impact of OFDI on GTFP diminishes as the institutional divergence between host and home countries increases. Building on this empirical foundation, we propose the following hypothesis:

H3.

Utilizing industrial robots stimulates GTFP improvement through FDI.

Fourth, by harnessing TI, using industrial robots can facilitate HQED. GTFP enhancement due to the advancement in technology-oriented industrial setups is evident in the reinforcement of the industrial structure driven by TI. According to the theory of structural effects, the economic composition of an area shifts from agricultural dominance to one marked by energy-intensive heavy industries, environmentally friendly services, and knowledge-intensive sectors [33]. The transformation of dynamic mechanisms that govern economic growth lies at the core of HQED. This scenario requires incorporating environmental factors into production functions to encapsulate the equilibrium between economic advancement and environmental preservation [34,35]. The mechanism supporting the widespread adoption of industrial robots, which contributes to HQED, can be attributed to multiple factors. For example, industrial robots symbolize technological advancement and represent a fundamental catalyst of economic growth [1]. Conversely, [36] assesses the influence of robot usage on economic growth using a neoclassical economic growth model. The model posits that robots and labor may be substitutes or complementary elements in varying degrees across occupations. Furthermore, it assumes that investments in robots can increase in response to shifts in demand. Consequently, the widespread deployment of robots may significantly foster economic growth, which catalyzes TI within enterprises [37] and elevates the knowledge level of businesses. According to science and technology, the major difference between developed and developing nations is that developed ones are always at the forefront of science and technology due to strong independent R&D capacities, which support technological advancement and fuel sustained economic growth and successful HQED. For developing ones, independent innovation is lacking, and technical advancement is mainly obtained by introducing new technology, which plays a minor role in encouraging HQED. Thus, we propose that

H4.

Utilizing industrial robots stimulates GTFP improvement through TI.

Fifth, IRAs facilitate HQED through R&D. In recent years, increasing concerns regarding environmental sustainability drew attention to the role of R&D endeavors in bolstering GTFP. Multiple findings stress the impact of R&D activities in propelling green production methods and sustainable development [9,38,39]. Numerous studies examine the influence of independent R&D technology spillovers on GTFP by considering diverse technology sources. For example, [40] imply that R&D augments productivity within analogous industries. [41] explore how R&D investments enhance productivity by utilizing panel data. In summary, international research highlights the significant impact of R&D activities on GTFP. Through avenues, such as environmental innovation, the development of clean energy technologies, and policy backing, R&D plays a pivotal role in driving the shift toward green production methods, which enhances GTFP and facilitates synergy between economic growth and environmental sustainability. Therefore, we posit that

H5.

Utilizing industrial robots stimulates improvement in GTFP through R&D.

2.5. Research Gaps and Theoretical Contributions

While substantial evidence supports the role of automation, innovation, and regulation in shaping productivity and sustainability, the literature remains fragmented. Most existing studies examine these factors in isolation, failing to explore their combined and interactive effects. Moreover, few works have explicitly linked industrial robot applications with GTFP, particularly in a cross-national context that accounts for development heterogeneity and regulatory complexity.

The theoretical contribution of this study lies in bridging these gaps. By integrating automation, environmental regulation, and innovation mechanisms into a single analytical framework, this research offers a more comprehensive understanding of how IRAs contribute to HQED. It also distinguishes between developed and developing economies, recognizing their differing capacities to adopt and benefit from robotic technologies.

Furthermore, this study addresses the threshold effects of ER; the mediating roles of FDI, TI, and R&D; and the nonlinear dynamics underlying automation’s impact. This interdisciplinary approach enriches both the automation and sustainability literature and provides actionable insights for policymakers aiming to balance economic modernization with environmental stewardship.

To synthesize the above theoretical perspectives and proposed hypotheses, we present a conceptual framework that captures the dynamic and multifaceted relationships among industrial robot adoption, innovation pathways, environmental regulation, and high-quality economic development. This framework is illustrated in Figure 2. This model serves as the foundation for the empirical strategy developed in the following sections.

3. Methodology

This study employs a three-part research design to examine whether industrial robot applications (IRAs) contribute to high-quality economic development (HQED) and whether threshold effects exist in this relationship. First, the slacks-based measure (SBM) approach is used to compute green total factor productivity (GTFP) as a key indicator of HQED. Next, we employ fixed-effects models and the system-generalized method of moments (SYS-GMM) to assess cross-group heterogeneity in the impact of IRAs on HQED, distinguishing between developed and developing economies.

To ensure the robustness of our findings, we conduct sensitivity analyses by substituting explanatory variables and alternative model specifications. Additionally, we apply the panel threshold model to identify potential threshold effects in the relationship between IRAs and HQED. Finally, we test the proposed hypotheses by examining the mediating effects of key economic factors, further exploring the mechanisms through which IRAs influence HQED.

3.1. Model Selection

3.1.1. GTFP Measurement

Green total factor productivity (GTFP) is a critical measure of economic efficiency and environmental sustainability, capturing both production performance and pollution control. This study employs data envelopment analysis (DEA) to assess GTFP, utilizing the slacks-based measure (SBM)–global Malmquist–Luenberger (GML) index and the directional distance function (DDF)–GML model. As a nonparametric approach, DEA facilitates the evaluation of multiple input–output systems, enabling a comprehensive assessment of technical efficiency [42]. It provides an objective measure of production efficiency while incorporating environmental constraints [43].

Since GTFP consists of desirable (expected) and undesirable (unexpected) outputs, we employ the SBM model to analyze the interrelationships among input utilization, production efficiency, and environmental impact. The SBM model effectively addresses inefficiencies by considering slack variables in the production process [44]. Compared to conventional linear programming models, SBM is both global and transitive, ensuring a feasible solution when evaluating efficiency [20]. The SBM model is defined as follows:

(4) $\begin{matrix} m i n ρ = \frac{1 - \frac{1}{m} \sum_{i = 1}^{m} S_{i}^{-} / x_{i k}}{1 + \frac{1}{q 1 + q^{2}} (\sum_{r = 1}^{q^{2}} S_{r}^{+} / y_{r k} + \sum_{t = 1}^{q^{2}} S_{t}^{b} / b_{r k})} \\ X λ + S^{-} = x_{k} \\ Y λ + S^{+} = y_{k} \\ B λ + S^{b} = b_{k} \\ λ, S^{-}, S^{+} \geq 0 \end{matrix}$

where

m

represents the number of input elements,

q_{1}

denotes the expected output classes, and

q_{2}

represents the unexpected output types. The variables

x_{i k}

y_{r k}

, and

b_{r k}

indicate the input, expected output, and unexpected output components of GTFP, respectively. Slack variables

{S i}^{-}

{S r}^{+}

, and

{S_{t}}^{b}

capture inefficiencies in input utilization, desirable output, and undesirable output. The efficiency score derived from this model provides an indicator of green productivity performance, where a higher efficiency value signifies superior economic and environmental performance.

Using the SBM model, we construct the GML index, which measures changes in green efficiency [20] over time. The GML index (Formula (5)) is given by

(5) $G M L_{i}^{i + 1} = \frac{1 + \bar{D_{0}^{G}} (x^{i}, y^{i}, b^{i} : y^{i}, - b^{i})}{1 + \bar{D_{0}^{C}} (x^{i + 1}, y^{i + 1}, b^{i + 1} : y^{i + 1}, - b^{i + 1})} .$

When the rate of output increase is greater than those of factor inputs, then the GML productivity index calculates the percentage growth. A GTFP increase or decrease is implied when GML > 1 or <1, respectively. Lastly, no change occurs in GTFP when GML = 1. This study considers that all GTFPs in the baseline period are 1 following the literature [45,46]. The reason is that GTFP is an indicator of growth across months. To determine GTFP in the following period, GTFP in the base period is multiplied by the GML index of the subsequent period. GTFP for other years is calculated in a similar manner.

In Formula (5), $D_{0}^{G}$ denotes global DDF, $x$ is the input component pertaining to GTFP, and $y^{i}$ is the anticipated output component associated with GTFP. Furthermore, $y$ represents the unanticipated output element within the context of GTFP. Notably, $t$ holds temporal significance by corresponding to the present period, while $t + 1$ signifies the subsequent period.

For robustness testing, we replace the DDF-GML model to measure GTFP. To assess the efficacy of each DMU, we first combine the production possibility set and DDF. Formula (6) is a typical one for expressing DDF [47]:

(6) $\vec{D_{0}} (x, y, d, g) = s u p \{β : (y, d) + β g \in P (x)\} .$

Several key variables in Formula (7) assume pivotal roles: $x$ represents the input parameter and $y$ symbolizes a desirable output. Notably, $d$ denotes the undesirable output, which serves as a crucial aspect. Incorporating the direction vector $g$ into the equation, specifically as $g = (y, - d)$ , imparts a stringent requirement for a dual objective: to augment and curtail desirable and undesirable outputs, respectively. This dual objective characterizes the challenges and objectives inherent to this model. $P (x)$ includes the range of production possibilities under the purview of environmental technology operations. It characterizes feasible combinations of inputs and outputs given technological constraints. Furthermore, $G = (g_{y}, g_{d})$ signifies the direction vector that encompasses the extension of output possibilities. This vector, $G$ , plays a fundamental role in defining the direction and magnitude of expansion in output space. The following linear programming problem can be used to determine the DDFs (Formulas (7)):

(7) $\begin{matrix} \vec{D_{0}^{t}} (x^{t}, y^{t}, d^{t}; y^{t}, - d^{t}) = m a x β \\ s, t, \{\begin{matrix} \sum_{j = 1}^{J} z_{j}^{t} y_{j n} \geq (1 + β) y_{j^{'} n^{'}}^{t} & p = 1,2, \dots, P \\ \sum_{j = 1}^{J} z_{j}^{t} d_{j k}^{t} \geq (1 - β) d_{j^{'} k^{'}}^{t} & k = 1,2, \dots, K \\ \sum_{j = 1}^{J} z_{j}^{t} x_{j n}^{t} \leq x_{j^{'} n^{'}}^{t} & n = 1,2, \dots, N \\ z_{j}^{t} \geq 0, & j = 1,2, \dots, J, \end{matrix} \end{matrix}$

where

j

represents the number of DMUs,

p

denotes the number of desirable outputs,

k

is the number of undesirable outputs,

z_{j}^{t}

pertains to the weight of period

t

, and

β

is the largest proportion that causes the decrease and increase in undesirable and desirable outputs, respectively. Finally, using the DDF as a foundation, this study uses the GML productivity index to describe country GTFP according to Formula (5).

3.1.2. Fixed Effects and the System Generalized Method of Moments Model

Utilizing the model defined in Formula (8) [48], the current study investigates heterogeneity in the effects of IRAs on HQED across developing and developed nations:

(8) ${G T F P}_{i j} = α_{0} + α_{1} {I R A}_{i j} + α_{2} C_{i j}^{k} + μ_{i} + θ_{j} + ξ_{i j} .$

We utilize lagged levels as instruments and employ the SYS-GMM estimator developed by [49] and enhanced by [50]. This estimator combines a set of moment conditions derived from the equation in levels with the conventional set of moment conditions in first differences.

We employed the SYS-GMM method for estimation due to its capacity to effectively integrate data pertaining to the cross-individual variation in levels and the within-household variation in changes. This method enables the control of time-invariant household-specific effects and addresses endogeneity associated with lagged dependent variables. Additionally, we recommend SYS-GMM, because it can mitigate bias stemming from variable omission, unobserved household heterogeneity, measurement error, and endogeneity, which frequently impacts the dependent variable. Moreover, two-step SYS-GMM estimation demonstrates superiority over one-step estimation due to its resilience to heteroskedasticity, panel-specific autocorrelation, and the Windmeijer correction for finite samples, which mitigates standard error bias. Furthermore, the instrument was condensed to enhance outcome robustness and incorporated a latency constraint for level and transformed equations. SYS-GMM exhibits greater efficiency and resistance to heteroskedasticity, and autocorrelation compared with the one-step SYS-GMM method [51]. Consequently, the generalized two-step system moment regression technique emerges as the optimal method of analysis.

3.1.3. Panel Threshold Model

Grounded in theoretical analysis and H2, we propose that the influence of IRAs on HQED may exhibit interval-dependent effects and a nonlinear relationship dependent on the degree of the industrialization of industrial robots. To address this concern, we adopt the panel threshold regression model by [52] and implement [53] command. To alleviate potential bias associated with artificially imposed intervals, we use an endogenous interval segmentation strategy based on inherent data characteristics. We employ Formula (9) as the analytical model for exploring the relationship between industrial robot density and HQED over time:

(9) $\begin{matrix} {G T F P}_{i j} = α_{0} + α_{1} {I R A}_{i j} I ({l n E R}_{i j} < γ_{1}) \\ + α_{2} {I R A}_{i j} I (γ_{1} < {l n E R}_{i j} < γ_{2}) + \dots \\ + α_{n} {I R A}_{i j} I (γ_{n - 1} < l n {E R}_{i j} < γ_{n}) + \\ α_{n + 1} {I R A}_{i j} I (l n {E R}_{i j} > γ_{n}) + \\ \sum_{k} φ_{k} C_{i j}^{k} + μ_{i} + θ_{j} + ξ_{i j} \end{matrix}$

Following [54], we use GTFP as the explanatory variable and to measure HQED; the core explanatory variable is the stock of industrial robots per 10,000 people. The threshold variable is the logarithm of ERs, which is multiplied by 1 to obtain the natural logarithm. When the threshold variable ER is $l n E R : {l n E R}_{i j} < γ_{1}, γ_{n - 1} < {l n E R}_{i j} < γ_{n}, l n {E R}_{i j} > γ_{n}$ , then the regression coefficients of the effect of the core explanatory variable (IRAs) on the explanatory variable (GTFP) are $α_{1}, α_{2}$ $\dots α_{n}, α_{n + + 1}$ . Urbanization (U), foreign trade (FT), government interference (GI), and labor supply quality (LSQ) are examples of control variables in $X^{k}$ , which pertains to a set of schematic functions called $I (\cdot)$ . We calculate the natural logarithm of U, FT, and GI by multiplying each by 1. The value enclosed within parentheses takes on a value of 1 when the specified inequality is met; otherwise, it is $0$ . In this context, $u i$ is the fixed effect (FE) attributed to individual countries, while $ϑ_{t}$ represents those associated with each time point. These effects remain consistent across time and countries. Additionally, $φ_{k}$ signifies the regression coefficient linked to each control variable with the inclusion of a random error term in the equation.

3.1.4. Mediation Effect Model

To examine the intermediary mechanism through which IRAs impact HQED and to validate H3–H5, we used the mediation effect model by [55] and [56]. Although the stepwise method by [57] is a prevalent approach for testing mediating effects, others argue in favor of the direct assessment of coefficient products by bootstrapping. Consequently, we constructed the mediation effect model as outlined in Formulas (10) and (11).

(10) $M_{i j} = α_{0} + A \times {I R A}_{i j} + β^{'} C_{i j}^{k} + μ_{i} + θ_{j} + ξ_{i j}$

${G T F P}_{i j} = α_{0} + B \times M_{i j}$

(11) $D^{'} {I R A}_{i j} + β^{'} C_{i j}^{k} + μ_{i} + θ_{j} + ξ_{i j}$

where the intermediary variable M is composed of three constituent elements, namely, R&D, TI, and FDI. The symbols align with those featured in Formula (9). Analysis of the intermediate effects encompasses the following stages. First, we estimate Formula (9) to explore the impact of IRAs on HQED in developing and developed countries. Subsequently, if coefficients A, B, and C′ demonstrate statistical significance, and C′ falls below a particular threshold, then we infer the presence of a mediating effect when assessing Formula (10). To verify the existence of mediating effects, we conduct a series of Sobel tests if A or B lacks statistical significance [58].

3.2. Descriptive Statistics and Data Source

3.2.1. Descriptive Statistics

To evaluate high-quality economic development (HQED) across nations and facilitate cross-country comparisons, this study follows the approach of [21] and employs [59] to compute green total factor productivity (GTFP) using the SBM-GML model. To ensure robustness, we also apply the DDF-GML model, allowing for an alternative measure of GTFP reliability. GTFP is defined as the aggregate value of green production divided by the sum of green production input, reflecting efficiency in resource utilization and environmental performance. The core explanatory variable in this study is industrial robot applications (IRAs), measured by the number of industrial robots per 10,000 workers, denoted as IRA [54]. To further validate the robustness of our findings, we introduce a logarithmically transformed version of this variable, labeled IRA2. In examining the threshold effects of environmental regulations (ERs) on HQED, we adopt per-capita CO₂ emissions as a proxy [6]. ER is calculated by aggregating cumulative emissions from the base year $Y_{0}$ to year $Y_{i}$ , normalized by the population of the final year.

This approach accounts for the intergenerational responsibility of emissions, aligning with [60] framework of cumulative greenhouse gas emissions per capita. The formula for ER calculation is as follows:

(12) $E R (Y_{i}) = \frac{E R (Y_{i})}{P (Y_{i})} = \frac{[\sum_{Y = Y_{0}}^{Y_{i}} E R (Y)]}{P (Y_{i})}$

The impact of IRAs on GTFP growth operates through several key mechanism variables, including foreign direct investment (FDI), technological innovation (TI), and research and development (R&D) [54,61]. TI is quantified by the total number of patents granted within a country, while R&D is measured using the logarithmic transformation of R&D personnel per one million people.

To control for potential institutional, economic, and policy-related factors influencing HQED, we incorporate a set of control variables based on the Cobb–Douglas production function [62] and previous empirical studies [63,64,65]. These include urbanization (U), measured as the proportion of the urban population relative to the total population; labor structure quality (LSQ), defined as the ratio of the labor force to the total population; foreign trade (FT), calculated as the ratio of total imports and exports to GDP; and government intervention (GI), represented by the share of fiscal expenditure in GDP. To address potential nonlinearity and heteroscedasticity, we apply a logarithmic transformation to U, FT, and GI after adding 1.

The dataset used in this study is constrained by data availability from the International Federation of Robotics (IFR), which has only provided systematic data on industrial robots and the impact of COVID-19 since 1993. Given the prevalence of missing data in many developing countries, we select 16 developed and 16 developing nations as our sample, covering the period 1993 to 2019. Additionally, winsorized shrinkage is applied to key variables to mitigate the influence of extreme outliers. Table 2 presents the descriptive statistics for the variables used in this study.

3.2.2. Data Sources

The primary data sources are as follows. Data on IRAs were derived from the IFR, which compiles statistics on IRAs across diverse industries and countries. Data on industrial robot adoption and inventory across countries were plotted. Second, raw data were used to calculate U, LSQ, FT, GI, FDI, R&D, and ER. Data were taken from the database of the Word Development Indicators (2023) by the World Bank. Lastly, data on the degree of scientific and TI were sourced from the database of the World Intellectual Property Organization (2023). Furthermore, the exclusion of post-COVID-19 pandemic data is based on the following logical considerations.

To begin with, the elimination of the influence of exogenous shocks is a crucial factor. The COVID-19 pandemic imposed an unprecedented shock on the global economy, causing severe disruptions in supply chains, labor market instability, and large-scale fiscal and monetary interventions by governments. These factors are likely to distort the normal relationship between economic growth and technological adoption. As the primary objective of this study is to examine the structural impact of industrial robot adoption on economic growth and green total factor productivity (GTFP), it is necessary to exclude short-term exogenous influences by using pre-pandemic data. Second, maintaining consistency in panel data analysis is an essential consideration. The panel data method (Panel Data Analysis) employed in this study compares national data observed over a specific period. However, post-2020 data may significantly deviate from previous economic patterns. For instance, factory shutdowns, the widespread adoption of remote work, and structural changes in industries during the pandemic may have abnormally altered the relationship between industrial robot adoption and economic growth. To ensure the consistency and reliability of long-term panel data, this study includes data only up to 2019. Third, the issue of incomplete data during the COVID-19 period further supports this decision. Data from 2020 onward may suffer from accuracy issues in some countries. Particularly in developing economies, economic and environmental data collection was disrupted during the pandemic, leading to a lack of reliable information on key variables such as industrial robot adoption rates and foreign direct investment (FDI). Such data deficiencies could compromise the credibility of the analysis. Finally, the scope of data selection aligns with the research objectives of this study. The core aim is to analyze the structural effects of industrial robots on long-term economic growth and green productivity. Thus, it is more appropriate to base the analysis on relatively stable long-term data rather than including the post-pandemic period, which is characterized by heightened policy volatility and economic uncertainty.

4. Empirical Analysis

4.1. Heterogeneity Analysis Across Groups

Table 3 presents the results of the heterogeneity analysis comparing the effects of industrial robot applications (IRAs) on high-quality economic development (HQED) between developed and developing countries. The analysis is conducted using fixed-effects (FE) models to control for unobserved heterogeneity across nations.

The findings confirm that IRAs play a significant role in fostering HQED, with positive regression coefficients observed across both groups. However, a notable distinction emerges between developed and developing economies. While the impact of IRAs on HQED is statistically significant in both groups, the regression coefficients indicate that IRAs exert a stronger influence in developing countries compared to developed ones. This suggests that IRAs contribute more substantially to productivity growth and structural transformation in economies where industrial automation adoption is still evolving. These results highlight the crucial role of IRAs in shaping HQED and provide empirical support for H1, confirming that industrial automation positively influences sustainable economic development.

4.2. Exploring the Threshold Effect of IRAs on ERs in Relation to GTFP Enhancement

Initially, we employed the [52] bootstrap method to identify the presence of the threshold effect. Table 4 illustrates the outcomes, which informed the specific form of the panel threshold model and indicates that the single and double thresholds for industrial robots pass the 1% significance test, while the triple threshold test yields nonsignificant results. Consequently, IRAs exhibit a double-threshold effect with ER as the critical variable. Afterward, we conducted threshold regression analysis. Table 5 outlines the results, which clearly indicate that the double threshold for ERs, as represented by per-capita CO₂ emissions, is 2.7162 and 2.7914. Across the three intervals defined by these thresholds, the coefficient of IRAs on GTFP progressively diminishes. As Table 5 suggests, the parameter values of lnER for developing and developed countries are 2.97 and 2.137, respectively, which positions them on opposite sides of the second and first thresholds, respectively. This result aligns with H2. Notably, the regression results demonstrate that the promotional effect of IRAs on GTFP intensifies with the decrease in ER as the second threshold is crossed, which facilitates HQED. Thus, H2 is supported.

4.3. Robustness Test

4.3.1. Effect of IRAs on GTFP

To determine that a nation has obtained HQED, scholars utilize the relative value technique [66,67]. Thus, we use GTFP derived using the DDF-GML model as the explanatory variable. The stock of industrial robots per 10,000 workers is used to estimate Formula (3) as a model with the addition of 1 to the logarithm that serves as a stand-in for the explanatory variable. Formula (8) is re-estimated as a model in this part using the FE and SYS-GMM models. Table 6 displays the results after the explanatory variable and explained variable have been swapped out. In Table 6, AR(2) is >0.1, which suggests that the residual components of each equation exhibit second-order serial independence. Simultaneously, the Sargan statistics exhibits nonsignificance, which affirms the reliability of the instruments employed in the construction of each model variable. With regard to the coefficients, the outcomes derived from substituting the dependent and independent variables of the capture and intersection groups closely align with those from the preceding section, which underscores the robustness of the findings.

4.3.2. Threshold Model Robustness

Following the substitution of the explanatory and explained variables, the testing results of the existence of thresholds demonstrate that IRAs maintain a significant dual-threshold effect on GTFP with ERs serving as the threshold. Table 7 provides the estimation results. As ER exceeds the initial and secondary thresholds, its influence on GTFP growth gradually wanes, but the regression results remain consistent. Additionally, when the explained variable is substituted with DDF-GML, the dual-threshold values for ER are 2.4996 and 2.8247, respectively. In this scenario, the lnER parameter values for developing and developed countries are recorded as 2.97 and 2.137, respectively, which places them inside and outside the second threshold, respectively. Analyzing the coefficients reveals a marked reduction in the contribution of industrial robots to GTFP growth after surpassing the second threshold, which is notably lower compared with the two other intervals. Nevertheless, the regression outcomes remain unaltered.

Nevertheless, endogeneity must be considered, which suggests that countries with advanced economies and high technological capacities opt for IRAs in production despite the varying impacts cited in the regression analysis. Industrial robots play a pivotal role in propelling economic growth. To address this concern, we construct instrumental variables based on the number of industrial robots adopted in other countries and duration since the inception of light industrial robots. This estimation is conducted using the 2SLS method.

In general, once the worldwide count of industrial robots reaches a threshold, and countries engage in competitive adoption, IRAs in one country do not significantly impact the HQED of other nations. To meet the core assumptions of correlation and exclusion, we employ an instrumental variable, which is represented as the global industrial robot inventory minus the industrial robot count of a country (lnWIR). Table 8 provides the detailed results. Regression outcomes indicate that the first-stage instrumental variable regression coefficient for the re-estimation of industrial robots is significantly positive for developed countries, whereas it is significantly negative for developing countries. This result implies a similarity between the global industrial robot inventory and that of developed nations with a positive correlation between the stock of industrial robots in developing countries and the first-stage instrumental variable in developed countries. This outcome agrees with those of [68]. The Kleibergen–Paap–Wald rk F-statistics and the Cragg–Donald Wald F-statistics surpass the critical value of 16.38, as established by the Stock–Yogo weak instrumental variable identification F-test at the 10% significance level. This finding underscores the rationality and efficacy of the instrumental variables, which confirms the absence of issues associated with weak instrumental variables. When applying the instrumental variables, the estimated results under column (2) reveal a significantly positive regression coefficient at the 1% level of significance. This aspect indicates that the adoption of industrial robots leads to a substantial and positively impactful contribution, which aligns with the baseline regression findings and affirms the robustness of the results.

We intend to locate an exogenous shock in the sample period (1993–2019) to retest the instrumental variables and manage endogeneity. This article, which adopts the method of [65], creates the instrumental variable post2006 Revolution year, where post2006 is a dummy variable that denotes the perception of the influence of light industrial robots. The timing of revolutions reveals regional variances and regional differences in industrial basis. Table 9 displays the regression results. The regression coefficient for the first stage is substantially positive at the 1% level, which is higher for developing than developed nations and suggests that nations during the Industrial Revolution deployed heavier robots in the subsequent years. The positive regression coefficient of the second stage demonstrates that light industrial robots boosted the GTFP of nations using IRAs, with emerging nations benefiting the most. The findings in Table 9 consider light industrial robots with endogeneity as an instrumental variable for analyzing the effects of technological shocks on GTFP.

4.4. Analysis of Mechanisms

The estimates suggest that IRAs significantly contributed to HQED. However, how is this possible? Based on previous assumptions, we aim to examine the detailed conduction process of FDI, TI, and R&D.

4.4.1. Mechanism of FDI

Scholars theoretically and empirically confirm the impact of FDI on IRAs and find that FDI is highly correlated with the national adoption of robots [32,69,70]. In addition, many studies examine the influence of FDI on GTFP growth [61,71,72]. To explore the mediating effect of FDI, scholars use two-step econometric and nonlinear mediating effect models for empirical research [73].

Table 10 presents the mediating effect of FDI. This study reveals that FDI is a mediator in the relationship between IRAs and HQED. Notably, the coefficient estimates for industrial robots under columns (2) and (5) exhibit a significant positive impact on FDI growth for both groups of nations. The results under columns (3) and (6) indicate that the IRA coefficients of developed countries do not significantly change after introducing FDI in the baseline model, whereas those of developing countries increase. Combining baseline model estimates under columns (1) and (4), the results illustrate that FDI does not exert a mediating effect on developed countries but plays a partial role in developing countries in IRAs to promote HQED. This finding depicts that IRAs will promote HQED through FDI; hence, H3 is supported.

4.4.2. Mechanism of TI

The foundational concept of modern growth theory, which posits that exogenous technological progress is the sole driver of economic growth, originates from the seminal economic growth model of [1]. Ref. [74] gauged regional innovation capabilities using the number of patent applications and investigated the linear and nonlinear effects of innovation on GTFP. Ref. [75] unveiled a notable negative impact of the interaction between TI and economic development level on GTFP [75]. Additionally, other scholars revealed a significant mediating role of green patents between interregional investment and the local GTFP of host cities (e.g., [76]).

Table 11 presents the estimation of the mechanism effect of TI, which highlights its influence on the impact of IRAs on HQED. Notably, the significantly positive estimated coefficients of IRAs under columns (2) and (4) indicate that IRAs foster TI. Moreover, the higher estimation value in the latter column implies that developing countries benefit more from the promotion of TI compared with their developed counterparts.

Within the domain of industrial robots, TI assumes a mechanistic role in advancing high-quality economic growth. Specifically, it accounts for approximately 42.15% and 0.30% of the total effects in developed and developing countries, respectively. Notably, the contribution of industrial robots to TI and the share of the mechanism effect of TI in the total effect are substantially higher in developed compared with developing countries. This observation underscores the capacity of industrial robots to stimulate GTFP growth through TI, which supports H4.

4.4.3. R&D Mechanism

Ref. [77] comprehensively analyzed 38 distinct industrial sectors in China and examined the realms of GTFP and its underlying determinants. The study yielded intriguing findings: notably, investments in R&D exerted a positive influence on GTFP growth within these industrial sectors. An interesting nuance emerged in which the promoting effect of R&D investment proved particularly evident in sectors characterized by high levels of emissions.

In this regard, ref. [78] examined Japanese manufacturing companies from 2001 to 2010 to disentangle the intricate relationships among corporate R&D investment, environmental performance, and financial outcome. The empirical findings demonstrated a surprisingly positive association between CO₂ emissions and green R&D projects. Understanding the influence of R&D investment orientations on GTFP in this context is crucial for solving resource depletion and environmental degradation. Thus, the authors empirically investigated the effects and underlying mechanisms of environmentally induced and traditional R&D in fostering GTFP within the provincial landscapes in China from 2004 to 2019 using the perpetual inventory method, meta-frontier DEA method, and the mediation effect test [53]. Thus, the current study aims to elucidate the complex effects and subtle mechanisms through which R&D endeavors promote GTFP.

Table 12 lists the estimated mechanism effect of R&D and demonstrates that R&D exerts no mechanism effect on the impact of industrial robots on HQED. Among them, the effect of the intermediary variable RD on the explained variable GTFP is nonsignificant (0.001, 0.002), which indicates no intermediary effect. This result determines if IRAs can help developing and developed countries increase GTFP growth through R&D; thus, H5 is unsupported.

In summary, the findings encompass the notable disparity in the contributions of IRAs to HEQD among distinct economic groups, such as developing and developed countries. It also includes the intricate dimensions of threshold effects and heterogeneity related to IRAs. These insights collectively contribute to HQED advancement within national economies. To address the potential challenge of endogeneity, we employ instrumental variables, specifically the number of robots within a given country and the introduction of light industrial robots in 2006. These instruments are employed to substantiate the robustness of the findings, which consistently affirms the positive relationship between IRAs and GTFP augmentation. This validation reinforces the notion that the adoption of industrial robots plays a contributory role in nurturing HQED. Furthermore, empirical analysis elucidates the intricate mediating mechanisms, which underscores the pivotal roles of FDI and TI in channeling the effects of IRAs. These nuanced revelations represent novel contributions to the existing literature, which enriches the current understanding of the multifaceted interactions between industrial robots and the pathways to HQED.

This study aimed to determine if IRAs contribute to HQED advancement in developing and developed nations. Although the prior literature examined the interplay among artificial intelligence, the digital economy, economic development, and HQED, a notable research gap exists in the discourse of the simultaneous consideration of industrial robot adoption and its implications for HQED within a unified analytical framework. This study represents a pioneering effort to amalgamate IRAs and HQED within a cohesive framework to address this long-term research gap. This study significantly contributes to the field of HQED and offers essential references by conducting a thorough exploration of theoretical and empirical connections. Consequently, the findings potentially offer vital theoretical underpinnings and guidance for enhancing HQED in diverse contexts.

5. Conclusions and Implications

5.1. Conclusions

Using the classification standards by [79] and the [11], we employ the GDP grouping criteria to categorize countries into developing and developed economies. Based on this classification, we empirically examine the role of industrial robot applications (IRAs) in fostering high-quality economic development (HQED) by utilizing panel data on 32 countries from 1993 to 2019. The results yield several notable findings.

First, IRAs exhibit a statistically significant positive influence on HQED, as measured by green total factor productivity (GTFP). This confirms the theoretical expectation that automation can enhance production efficiency while contributing to environmental performance. Second, by incorporating environmental regulation (ER) as a threshold variable, this study reveals a dual-threshold effect. Specifically, the contribution of IRAs to GTFP diminishes after ER exceeds two identified threshold levels, indicating that the marginal benefits of IRAs may decline in highly polluted or stringently regulated environments. This nonlinear relationship suggests that while IRAs can effectively promote green productivity, their effectiveness is context-dependent and sensitive to regulatory saturation.

Third, the study identifies two key mediating mechanisms: foreign direct investment (FDI) and technological innovation (TI). IRAs indirectly enhance GTFP by facilitating FDI in developing countries, where external capital and technology transfer play critical roles in modernization. Similarly, in both developing and developed nations, IRAs stimulate technological innovation, thereby reinforcing green growth pathways. However, the study does not find strong empirical support for research and development (R&D) as a significant mediating variable, particularly in resource-constrained or institutionally underdeveloped regions.

Despite these insights, existing studies remain limited in several key respects. Most prior research treats industrial automation and environmental sustainability as isolated fields, lacking integrated empirical frameworks that capture their interaction [80]. Furthermore, the heterogeneous effects of IRAs across different development stages and environmental policy regimes are often neglected. Very few studies explore how threshold effects or mediating channels operate in a global comparative context using long-term panel data.

By systematically addressing these gaps, this study contributes to the literature in three ways. Academically, it bridges the disciplinary divide between automation economics and environmental policy, offering a unified framework to examine their intersection. Methodologically, it applies advanced threshold and mediation models to identify nonlinear and indirect effects, which are often overlooked. Practically, the findings offer actionable implications for policymakers aiming to balance technological upgrading with environmental governance, especially in emerging economies undergoing industrial transformation.

In conclusion, the empirical evidence confirms that IRAs play a vital role in promoting HQED, particularly through FDI and TI pathways. However, their effectiveness is contingent upon the regulatory environment and the stage of economic development. This nuanced understanding underscores the need for differentiated and adaptive policy approaches that harness automation for sustainable growth.

5.2. Implications

This study formulates comprehensive and differentiated policy recommendations aimed at leveraging industrial robot applications (IRAs) to foster high-quality economic development (HQED), grounded in the empirical heterogeneity across developing and developed countries.

First, both developing and developed economies should actively promote the adoption and integration of IRAs to accelerate intelligent industrial transformation. In general, industrial robot deployment can enhance production efficiency, reduce operational costs, and facilitate low-carbon development by minimizing energy usage and pollution emissions. However, the pathways and priorities differ by country type:

5.2.1. In Developing Countries

Governments should strategically guide industrial upgrading by attracting green technology-oriented foreign direct investment (FDI). Targeted investment incentives, such as tax relief or green zone policies, should be designed to steer FDI into high-value-added, environmentally responsible sectors rather than traditional low-tech industries. This approach will maximize technology spillovers, foster local innovation, and promote cleaner production methods.

5.2.2. In Developed Countries

The policy focus should shift toward enhancing domestic breakthrough innovation capabilities. Support for advanced R&D and deep-tech innovation—particularly in AI-driven robotics, smart manufacturing systems, and green industrial automation—will be essential to maintaining competitiveness and driving the next wave of productivity and environmental gains.

Second, differentiated capacity-building systems must be established to support the deployment and effective utilization of IRAs:

5.2.3. For Developing Nations

Vocational training programs should prioritize practical technical competencies, from robot operation to basic maintenance and troubleshooting. This will help bridge skill gaps and ensure that imported technologies are absorbed and applied effectively.

5.2.4. For Developed Nations

Policies should emphasize the cultivation of interdisciplinary innovation talent—engineers, data scientists, and systems integrators—who can lead the development and scaling of complex, AI-integrated automation systems. Collaborative initiatives between universities, enterprises, and government agencies should be incentivized to build innovation-driven ecosystems.

Third, governments should intensify efforts to integrate IRAs into broader sustainability agendas through enhanced environmental regulations (ERs). Although this study identifies a threshold effect in ERs—where overly stringent policies may reduce marginal benefits—well-calibrated regulations remain essential for incentivizing green transformation. Establishing clear environmental standards for automated production processes, and pairing them with financial incentives (such as subsidies for green machinery, carbon credits, or green certification), can drive the adoption of cleaner robot technologies. Governments may also develop public–private partnerships to fund joint research on next-generation green automation solutions, thereby promoting inclusive and long-term innovation.

Fourth, cross-border cooperation should be deepened to reduce technological inequality and foster shared innovation. International platforms for collaborative R&D in robotics, environmental technology, and smart manufacturing should be promoted, especially to enable knowledge transfer to developing countries. These platforms can facilitate standard harmonization, joint innovation programs, and multilateral funding initiatives.

Fifth, awareness-building and environmental education campaigns should be tailored to each development stage. For developing economies, increasing investor and enterprise understanding of the value of green technology and sustainable industrial practices is key. For developed nations, public discourse should focus on ethical automation, just transition strategies, and long-term competitiveness in green tech.

Analyzing the impact of IRAs on HQED across countries is inherently complex. This study reveals meaningful heterogeneity in the effects of IRAs, mediated by FDI and TI channels. However, data limitations—especially during the COVID-19 period—have restricted the inclusion of certain dynamic policy variables. Future research should incorporate richer, longitudinal datasets and apply quasi-experimental methods such as difference-in-differences (DID) and propensity score matching (PSM) to capture pre- and post-policy shifts more rigorously. Additionally, extending analysis to include other emerging automation technologies (e.g., collaborative robots, AI-embedded platforms) would offer a broader understanding of industrial modernization in the green era.

Author Contributions

Conceptualization, B.S., Y.-C.L. and S.-D.P.; methodology, Y.-C.L. and B.S.; software, Y.-C.L.; validation, B.S.; formal analysis, Y.-C.L.; investigation, S.-D.P.; data curation, Y.-C.L.; writing—original draft preparation, B.S., Y.-C.L.; writing—review and editing, B.S. and S.-D.P.; visualization, Y.-C.L.; supervision, S.-D.P. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data presented in this study are available on request from the corresponding author.

Conflicts of Interest

The authors declare no conflict of interest.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures and Tables

Figure 1 Association between GTFP and industrial robots from 1993 to 2019.

Figure 2 Conceptual framework of this study. The model investigates the effect of industrial robot density on green total factor productivity (HQED proxy), mediated by FDI and technological innovation, and moderated by environmental regulation. SYS-GMM and threshold models are applied for estimation and robustness testing.

Table 1

Subgroups of developing and developed countries.

Developed Countries	Developing Countries
America	Argentina
Austria	Brazil
Belgium	Bulgaria
Canada	China
Denmark	India
England	Indonesia
Finland	Malaysia
France	Mexico
Germany	The Philippines
Italy	Poland
Japan	Romania
The Netherlands	Russia
Norway	South Africa
South Korea	Turkey
Sweden	Ukraine
Switzerland	Vietnam

Data source: https://www.rug.nl/ggdc/productivity/pwt/?lang=en (accessed on 20 September 2024).

Table 2

Descriptive statistics.

Variables	Variable Name		Developing Countries					Developed Countries
Variables	Variable Name	Obs	Mean	SD	Min	Max	Obs	Mean	SD	Min	Max
Explanatory variable	GTFP(SBM-GML)	496	1.421	0.272	0.988	2.232	496	1.361	0.455	0.617	3.553
Explanatory variable	GTFP(DDF-GML)	496	1.047	0.2	0.727	1.643	496	1	0.334	0.454	2.616
Core explanatory variables	IRA	496	3.259	1.148	0.962	7.037	496	1.549	0.774	0.177	6.296
Core explanatory variables	lnIRA2	496	7.339	2.78	2.292	16.76	496	3.843	1.411	0.421	9.957
Threshold variables	lnER	496	2.97	0.155	2.5	3.318	496	2.137	0.556	−1.077	2.982
Mechanism variables	lnTI	496	10.147	3.463	6.425	28.585	496	8.927	2.088	4.292	23.287
	lnR&D	496	8.208	0.427	6.909	9.172	496	5.924	1.444	0	8.242
	FDI	496	4.274	8.662	−36.14	86.479	496	3.019	2.648	−2.757	31.228
Control variables	InU	496	4.389	0.083	4.202	4.585	496	3.95	0.47	1.99	4.522
	LSQ	496	0.601	0.134	0.363	1.031	496	0.497	0.154	0.086	1.186
	InGI	496	3.444	0.359	2.448	3.898	496	3.019	0.406	1.957	3.866
	lnFT	496	4.899	1.175	2.706	9.519	496	2.713	1.695	0.341	9.558

Table 3

Results of heterogeneity evaluation between the two groups.

Explanatory Variable	GTFP (SBM-GML)	GTFP (SBM-GML)
	Developing countries	Developed countries
	FE	FE
	(1)	(3)
IRA	0.355 ***(0.0029)	0.0400 *(0.0213)
Control variables	YES	YES
Country FE	YES	YES
Year FE	YES	YES
_cons	0.0582 **(0.0248)	0.580(0.695)
N	496	496
R-squared	0.997	0.994

Note: Values in parentheses are robust standard errors. *, **, and *** represent significance at the 10%, 5%, and 1%, levels, respectively.

Table 4

Results of threshold testing.

Model	F-Value	p-Value	BS-Times	1% Threshold	5% Threshold	10% Threshold
Single threshold	199.30 ***	0.0000	300	108.6929	70.7116	57.9210
Double threshold	96.61 **	0.0033	300	92.9981	61.9236	53.2354
Triple threshold	41.22	0.3633	300	91.3601	76.0178	66.7466

Note: ** and *** represent significance at the 10%, 5%, and 1% levels, respectively.

Table 5

Results of threshold effects.

Parameter Value	Estimated Value	T-Statistics	Confidence Interval
α1: lnER < 2.7162	0.233 ***	4.35	(0.1235174, 0.3416558)
α2: 2.7162 ≤ lnER < 2.7914	0.203 ***	3.97	(0.0988515, 0.3074735)
α3: lnER ≥ 2.7914	0.173 ***	3.59	(0.0748308, 0.2716971)
_cons	−0.721 **
Control variables	YES
Country FE	YES
Year FE	YES
N	992
R-squared	0.956

Note: ** and *** represent significance at the 10%, 5%, and 1% levels, respectively. Numbers in parentheses are robust standard errors. In the threshold significance test, the number of bootstraps was 300, and the number of seeds was 101.

Table 6

Robustness tests of the fixed effects and systematic generalized method of moments models across categories.

	Developing Countries				Developed Countries
	GTFP
	SBM-GML		DDF-GML		SBM-GML		DDF-GML
	FE	SYS-GMM	FE	SYS-GMM	FE	SYS-GMM	FE	SYS-GMM
IRA	0.172 **(0.0636)	0.387 ***(0.0183)			0.0296 *(0.0157)	0.234 ***(0.0134)
lnIRA2			0.184 ***(0.0232)	0.335 ***(0.0105)			0.0165 *(0.0091)	0.134 ***(0.0074)
Control variables	YES	YES	YES	YES	YES	YES	YES	YES
Country FE	YES	YES	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES	YES	YES
_cons	0.941(0.549)	0.149 ***(0.0568)	0.637(0.390)	0.0039(0.047)	0.427(0.509)	−0.0716(0.118)	0.632(0.685)	−0.0371(0.159)
AR(1)		0.000		0.000		0.000		0.000
AR(2)		0.184		0.923		0.501		0.546
Sargan’s statistics		0.166		0.201		0.140		0.138
Wald’s statistics		0.000		0.000		0.000		0.000
N	496	464	496	464	496	464	496	464
R-squared	0.805		0.983		0.994		0.994

Note: Numbers in parenthesis are robust standard errors. *, **, and *** represent significance at the 10%, 5%, and 1% levels, respectively.

Table 7

Robustness tests of the threshold model.

Parameter Value	GTFP (SBM-GML)			GTFP (DDF-GML)
Parameter Value	Estimated Value	Confidence Interval	Parameter Value	Estimated Value	Confidence Interval
α1: lnER < 2.7693	0.0979 ***(0.0298)	(0.0371883, 0.1585637)	α1: lnER < 2.4996	0.178 ***(0.0492)	(0.0775101, 0.2783952)
α2: 2.7693 ≤ lnER < 2.8581	0.0838 ***(0.0271)	(0.0284779, 0.1392063)	α2: 2.4996 ≤ lnER < 2.8247	0.119 **(0.0466)	(0.0239767, 0.2141756)
α3: lnER ≥ 2.8581	0.0741 ***(0.0254)	(0.0223507, 0.125781)	α3: lnER ≥ 2.8247	0.0965 **(0.0390)	(0.0169305, 0.1761031)
_cons	−0.606(0.470)		_cons	0.129(0.394)
Control variables	YES		Control variables	YES
Country FE	YES		Country FE	YES
Year FE	YES		Year FE	YES
N	992		N	992
R-squared	0.947		R-squared	0.932

Table 8

Instrumental variable: industrial robot application across groups-2SLS.

	Developing Countries		Developed Countries
	(1)	(2)	(3)	(4)
	First-Stage	Second-Stage	First-Stage	Second-Stage
	IRA	GTFP (SBM-GML)	IRA	GTFP (SBM-GML)
IRA	−0.1741 ***(0.0348)		0.1900 ***(0.0372)
lnWIR		0.330 ***(0.0161)		0.0896 ***(0.0158)
Control variables	YES	YES	YES	YES
Country FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
N	496	496	496	496
R-squared		0.998		0.993
First-stage F	185.13		357.03
p-Value	[0.000]		[0.000]
K-P-Wald rk F statistic		26.267		27.408
C-D Wald F statistic		31.568		19.204
Stock–Yogo weak ID test Critical values:10% maximal IV		16.38		16.38

Note: p-Values are in square brackets and *** represent significance at the 1%.

Table 9

Instrumental variable: utilization of light industrial robots (2006)-2SLS.

	Developing Countries		Developed Countries
	(1)	(2)	(3)	(4)
	First-Stage	Second-Stage	First-Stage	Second-Stage
	IRA	GTFP (SBM-GML)	IRA	GTFP (SBM-GML)
IRA	−0.1986 ***(0.0423)		−0.145 ***(0.0319)
post2006 × Revolution year		0.312 ***(0.0251)		0.0796 ***(0.0219)
Control variables	YES	YES	YES	YES
Country FE	YES	YES	YES	YES
Year FE	YES	YES	YES	YES
N	496	496	496	496
R-squared		0.998		0.994
First-stage F	237.01		281.02
p-Value	[0.000]		[0.000]
K-P-W rk F statistic		23.083		21.568
C-D Wald F statistic		37.147		37.960
Stock–Yogo weak ID test Critical values:10% maximal IV		16.38		16.38

Note: p-Values are in square brackets and *** represent significance at the 1%.

Table 10

Intermediary role of FDI.

	Developing Countries			Developed Countries
	(1)	(2)	(3)	(4)	(5)	(6)
	GTFP(SBM-GML)	FDI	GTFP(SBM-GML)	GTFP(SBM-GML)	FDI	GTFP(SBM-GML)
IRA	0.355 ***	1.423 ***	0.356 ***	0.040 *	1.002 **	0.040 *
	(0.003)	(0.314)	(0.003)	(0.021)	(0.427)	(0.021)
FDI			0.001 **			0.0002 *
			(0.0003)			(0.0001)
Control variables	YES	YES	YES	YES	YES	YES
Country FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
_cons	0.058 **	7.057	0.053 **	0.580	−312.631 **	0.515
	(0.025)	(6.748)	(0.023)	(0.695)	(141.033)	(0.696)
N	496	496	496	496	496	496
R-squared	0.997	0.195	0.997	0.994	0.192	0.994

Note: *, ** and *** represent significance at the 10%, 5%, and 1% levels, respectively.

Table 11

Intermediary role of TI.

	Developing Countries			Developed Countries
	(1)	(2)	(3)	(4)	(5)	(6)
	GTFP(SBM-GML)	lnTI	GTFP(SBM-GML)	GTFP(SBM-GML)	lnTI	GTFP(SBM-GML)
IRA	0.355 ***	1.486 **	0.351 ***	0.040 *	5.620 *	0.060 **
	(0.003)	(0.667)	(0.002)	(0.021)	(2.901)	(0.024)
lnTI			0.002 *			0.003 *
			(0.001)			(0.002)
Control variables	YES	YES	YES	YES	YES	YES
Country FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
_cons	0.058 **	−4.019	0.068 *	0.580	35.284	0.703
	(0.025)	(6.994)	(0.038)	(0.695)	(32.915)	(0.661)
N	496	496	496	496	496	496
R-squared	0.997	0.575	0.997	0.994	0.501	0.995

Note: *, ** and *** represent significance at the 10%, 5%, and 1% levels, respectively.

Table 12

Intermediary role of research and development.

	Developing Countries			Developed Countries
	(1)	(2)	(3)	(4)	(5)	(6)
	GTFP(SBM-GML)	lnR&D	GTFP(SBM-GML)	GTFP(SBM-GML)	lnR&D	GTFP(SBM-GML)
IRA	0.355 ***	−0.082	0.355 ***	0.040 *	0.182	0.036 *
	(0.003)	(0.261)	(0.003)	(0.021)	(0.111)	(0.019)
lnR&D			0.001			0.022
			(0.001)			(0.018)
Control variables	YES	YES	YES	YES	YES	YES
Country FE	YES	YES	YES	YES	YES	YES
Year FE	YES	YES	YES	YES	YES	YES
_cons	0.058 **	1.229	0.057 *	0.580	8.175 **	0.400
	(0.025)	(7.076)	(0.029)	(0.695)	(2.914)	(0.707)
N	496	496	496	496	496	496
R-squared	0.997	0.429	0.997	0.994	0.748	0.994

Note: *, ** and *** represent significance at the 10%, 5%, and 1% levels, respectively.

References

1. Solow, R.M. Technical change and the aggregate production function. Rev. Econ. Stat.; 1957; 39, pp. 312-320. [DOI: https://dx.doi.org/10.2307/1926047]

2. Epping, K.; Zhang, H. A Sustainable Decision-Making Framework for Transitioning to Robotic Welding for Small and Medium Manufacturers. Sustainability; 2018; 10, 3651. [DOI: https://dx.doi.org/10.3390/su10103651]

3. Yu, L.; Wang, Y.; Wei, X.; Zeng, C. Towards Low-Carbon Development: The Role of Industrial Robots in Decarbonization in Chinese Cities. J. Environ. Manag.; 2023; 330, 117216. [DOI: https://dx.doi.org/10.1016/j.jenvman.2023.117216] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36621317]

4. Acemoglu, D.; Restrepo, P. Artificial Intelligence, Automation, and Work. The Economics of Artificial Intelligence: An Agenda; Agrawal, A.; Gans, J.; Goldfarb, A. University of Chicago Press: Chicago, IL, USA, 2018; pp. 197-236.

5. Liu, B.; Yang, X.; Zhang, J. Nonlinear Effect of Industrial Robot Applications on Carbon Emissions: Evidence from China. Environ. Impact Assess. Rev.; 2024; 104, 107297. [DOI: https://dx.doi.org/10.1016/j.eiar.2023.107297]

6. Zhou, Q.; Zhang, X.; Shao, Q.; Wang, X. The Non-Linear Effect of Environmental Regulation on Haze Pollution: Empirical Evidence for 277 Chinese Cities during 2002–2010. J. Environ. Manag.; 2019; 248, 109274. [DOI: https://dx.doi.org/10.1016/j.jenvman.2019.109274] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/31374433]

7. Harding, T.; Smarzynska Javorcik, B. A Touch of Sophistication: FDI and Unit Values of Exports. World Bank Policy Research Working Paper; No. 4731 2009; Available online: https://openknowledge.worldbank.org/handle/10986/4087 (accessed on 5 June 2025).

8. Qiu, S.; Wang, Z.; Geng, S. How Do Environmental Regulation and Foreign Investment Behavior Affect Green Productivity Growth in the Industrial Sector? An Empirical Test Based on Chinese Provincial Panel Data. J. Environ. Manag.; 2021; 287, 112282. [DOI: https://dx.doi.org/10.1016/j.jenvman.2021.112282] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33711666]

9. Chen, X.; Nordhaus, W.D. Using Luminosity Data as a Proxy for Economic Statistics. Proc. Natl. Acad. Sci. USA; 2011; 108, pp. 8589-8594. [DOI: https://dx.doi.org/10.1073/pnas.1017031108]

10. Lyu, Y.; Wang, W.; Wu, Y.; Zhang, J. How Does Digital Economy Affect Green Total Factor Productivity? Evidence from China. Sci. Total Environ.; 2023; 857, 159428. [DOI: https://dx.doi.org/10.1016/j.scitotenv.2022.159428]

11. International Monetary Fund (IMF). IMF Data Portal. 2025; Available online: https://data.imf.org/ (accessed on 7 June 2025).

12. Fang, C.; Cheng, J.; Zhu, Y.; Chen, J.; Peng, X. Green Total Factor Productivity of Extractive Industries in China: An Explanation from Technology Heterogeneity. Resour. Policy; 2021; 70, 101933. [DOI: https://dx.doi.org/10.1016/j.resourpol.2020.101933]

13. Wang, S.; Zheng, Y.; Yang, H. Digital Economy and Green Total Factor Productivity in China: Empirical Evidence from Provincial Panel Data. PLoS ONE; 2024; 19, e0299716. [DOI: https://dx.doi.org/10.1371/journal.pone.0299716]

14. Rusiawan, W.; Tjiptoherijanto, P.; Suganda, E.; Darmajanti, L. Assessment of Green Total Factor Productivity Impact on Sustainable Indonesia Productivity Growth. Procedia Environ. Sci.; 2015; 28, pp. 493-501. [DOI: https://dx.doi.org/10.1016/j.proenv.2015.07.059]

15. Li, H.; Chen, C.; Umair, M. Green Finance, Enterprise Energy Efficiency, and Green Total Factor Productivity: Evidence from China. Sustainability; 2023; 15, 11065. [DOI: https://dx.doi.org/10.3390/su151411065]

16. Akisik, O. The Impact of Financial Development, IFRS, and Rule of Law on Foreign Investments: A Cross-Country Analysis. Int. Rev. Econ. Finance; 2020; 69, pp. 815-838. [DOI: https://dx.doi.org/10.1016/j.iref.2020.06.015]

17. Dritsaki, M.; Dritsaki, C. Trade Openness and Economic Growth: A Panel Data Analysis of Baltic Countries. Asian Econ. Financ. Rev.; 2020; 10, pp. 313-324. [DOI: https://dx.doi.org/10.18488/journal.aefr.2020.103.313.324]

18. Singh, S.; Sharma, G.D.; Radulescu, M.; Balsalobre-Lorente, D.; Bansal, P. Do Natural Resources Impact Economic Growth: An Investigation of P5+1 Countries under Sustainable Management. Geosci. Front.; 2024; 15, 101595. [DOI: https://dx.doi.org/10.1016/j.gsf.2023.101595]

19. Huang, G.; He, L.Y.; Lin, X. Robot Adoption and Energy Performance: Evidence from Chinese Industrial Firms. Energy Econ.; 2022; 107, 105837. [DOI: https://dx.doi.org/10.1016/j.eneco.2022.105837]

20. Oh, D.H. A Global Malmquist-Luenberger Productivity Index. J. Prod. Anal.; 2010; 34, pp. 183-197. [DOI: https://dx.doi.org/10.1007/s11123-010-0178-y]

21. Liu, Z.; Xin, L. Has China’s Belt and Road Initiative Promoted Its Green Total Factor Productivity? Evidence from Primary Provinces along the Route. Energy Policy; 2019; 129, pp. 360-369. [DOI: https://dx.doi.org/10.1016/j.enpol.2019.02.045]

22. Acemoglu, D.; Restrepo, P. The Race between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment. Am. Econ. Rev.; 2018; 108, pp. 1488-1542. [DOI: https://dx.doi.org/10.1257/aer.20160696]

23. Aklin, M. Re-Exploring the Trade and Environment Nexus through the Diffusion of Pollution. Environ. Resour. Econ.; 2016; 64, pp. 663-682. [DOI: https://dx.doi.org/10.1007/s10640-015-9893-1]

24. Hancevic, P.I. Environmental Regulation and Productivity: The Case of Electricity Generation under the CAAA-1990. Energy Econ.; 2016; 60, pp. 131-143. [DOI: https://dx.doi.org/10.1016/j.eneco.2016.09.022]

25. Hoque, A.; Baug, T.; Dewangan, L.; Wang, K.; Liu, T.; Mondal, S. Influence of Mid-Infrared Galactic Bubble on Surroundings: A Case Study on IRAS 16489-4431. arXiv; 2024; arXiv: 2405.17993Available online: https://arxiv.org/abs/2405.17993 (accessed on 5 June 2025). [DOI: https://dx.doi.org/10.25518/0037-9565.11811]

26. Bulus, G.C.; Koc, S. The Effects of FDI and Government Expenditures on Environmental Pollution in Korea: The Pollution Haven Hypothesis Revisited. Environ. Sci. Pollut. Res.; 2021; 28, pp. 38238-38253. [DOI: https://dx.doi.org/10.1007/s11356-021-13462-z] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/33733418]

27. Batabyal, A. Environmental Policy in Developing Countries: A Dynamic Analysis. Rev. Dev. Econ.; 1998; 2, pp. 293-304. [DOI: https://dx.doi.org/10.1111/1467-9361.00043]

28. Au, C.C.; Henderson, J.V. Are Chinese Cities Too Small?. Rev. Econ. Stud.; 2006; 73, pp. 549-576. [DOI: https://dx.doi.org/10.1111/j.1467-937X.2006.00387.x]

29. Lin, B.; Du, R.; Dong, Z.; Jin, S.; Liu, W. The Impact of Foreign Direct Investment on the Productivity of the Chinese Forest Products Industry. For. Policy Econ.; 2020; 111, 102035. [DOI: https://dx.doi.org/10.1016/j.forpol.2019.102035]

30. Wang, M.; Zhu, C.; Wang, X.; Ntim, V.S.; Liu, X. Effect of Information and Communication Technology and Electricity Consumption on Green Total Factor Productivity. Appl. Energy; 2023; 347, 121366. [DOI: https://dx.doi.org/10.1016/j.apenergy.2023.121366]

31. Song, M.; Du, J.; Tan, K.H. Impact of Fiscal Decentralization on Green Total Factor Productivity. Int. J. Prod. Econ.; 2018; 205, pp. 359-367. [DOI: https://dx.doi.org/10.1016/j.ijpe.2018.09.019]

32. Li, X.; Xiao, L. The Impact of Urban Green Business Environment on FDI Quality and Its Driving Mechanism: Evidence from China. World Dev.; 2024; 175, 106494. [DOI: https://dx.doi.org/10.1016/j.worlddev.2023.106494]

33. Grossman, G.M.; Krueger, A.B. Economic Growth and the Environment. Q. J. Econ.; 1995; 110, pp. 353-377. [DOI: https://dx.doi.org/10.2307/2118443]

34. Feng, C.; Huang, J.B.; Wang, M. The Sustainability of China’s Metal Industries: Features, Challenges and Future Focuses. Resour. Policy; 2019; 60, pp. 215-224. [DOI: https://dx.doi.org/10.1016/j.resourpol.2018.12.006]

35. Zheng, X.; Yu, H.; Yang, L. Technology Imports, Independent Innovation, and China’s Green Economic Efficiency: An Analysis Based on Spatial and Mediating Effect. Environ. Sci. Pollut. Res.; 2022; 29, pp. 36170-36188. [DOI: https://dx.doi.org/10.1007/s11356-021-17499-y]

36. Hanson, R. Economic Growth Given Machine Intelligence; Technical Report University of California: Berkeley, CA, USA, 2001; Available online: http://hanson.gmu.edu/aigrow.pdf (accessed on 5 June 2025).

37. Agrawal, A.; Gans, J.; Goldfarb, A. Prediction, Judgment, and Complexity: A Theory of Decision-Making and Artificial Intelligence. The Economics of Artificial Intelligence: An Agenda; University of Chicago Press: Chicago, IL, USA, 2018; pp. 89-110.

38. Yang, X.; Jia, Z.; Yang, Z. How Does Technological Progress Impact Transportation Green Total Factor Productivity: A Spatial Econometric Perspective. Energy Rep.; 2021; 7, pp. 3935-3950. [DOI: https://dx.doi.org/10.1016/j.egyr.2021.06.078]

39. Wang, Y.; Bai, Y.; Quan, T.; Ran, R.; Hua, L. Influence and Effect of Industrial Agglomeration on Urban Green Total Factor Productivity: On the Regulatory Role of Innovation Agglomeration and Institutional Distance. Econ. Anal. Policy; 2023; 78, pp. 1158-1173. [DOI: https://dx.doi.org/10.1016/j.eap.2023.04.024]

40. Badinger, H.; Egger, P. Productivity Spillovers across Countries and Industries: New Evidence from OECD Countries. Oxf. Bull. Econ. Stat.; 2016; 78, pp. 501-521. [DOI: https://dx.doi.org/10.1111/obes.12122]

41. Bengoa, M.; Martínez-San Román, V.; Pérez, P. Do R&D Activities Matter for Productivity? A Regional Spatial Approach Assessing the Role of Human and Social Capital. Econ. Model.; 2017; 60, pp. 448-461. [DOI: https://dx.doi.org/10.1016/j.econmod.2016.09.005]

42. Sueyoshi, T.; Yuan, Y.; Goto, M. A Literature Study for DEA Applied to Energy and Environment. Energy Econ.; 2017; 62, pp. 104-124. [DOI: https://dx.doi.org/10.1016/j.eneco.2016.11.006]

43. Tyteca, D. On the Measurement of the Environmental Performance of Firms: A Literature Review and a Productive Efficiency Perspective. J. Environ. Manag.; 1996; 46, pp. 281-308. [DOI: https://dx.doi.org/10.1006/jema.1996.0022] Available online: https://www.researchgate.net/profile/Daniel-Tyteca/publication/223445514 (accessed on 5 June 2025).

44. Tone, K.; Tsutsui, M. Dynamic DEA: A Slacks-Based Measure Approach. Omega; 2010; 38, pp. 145-156. [DOI: https://dx.doi.org/10.1016/j.omega.2009.07.003]

45. Zhong, J.; Li, T. Impact of Financial Development and Its Spatial Spillover Effect on Green Total Factor Productivity: Evidence from 30 Provinces in China. Math. Probl. Eng.; 2020; 2020, 5741387. [DOI: https://dx.doi.org/10.1155/2020/5741387]

46. Xue, Y.; Liu, K. Regional Differences, Distribution Dynamics, and Convergence of Air Quality in Urban Agglomerations in China. Sustainability; 2022; 14, 7330. [DOI: https://dx.doi.org/10.3390/su14127330]

47. Chambers, R.G.; Chung, Y.; Färe, R. Profit, Directional Distance Functions, and Nerlovian Efficiency. J. Optim. Theory Appl.; 1998; 98, pp. 351-364. [DOI: https://dx.doi.org/10.1023/A:1022637501082]

48. Liu, H.; He, Q. The Effect of Basic Public Service on Urban-Rural Income Inequality: A Sys-GMM Approach. Econ. Res.-Ekon. Istraž.; 2019; 32, pp. 3205-3223. [DOI: https://dx.doi.org/10.1080/1331677X.2019.1661005]

49. Arellano, M.; Bover, O. Another Look at the Instrumental Variable Estimation of Error-Components Models. J. Econom.; 1995; 68, pp. 29-51. [DOI: https://dx.doi.org/10.1016/0304-4076(94)01642-D]

50. Blundell, R.; Bond, S. Initial Conditions and Moment Restrictions in Dynamic Panel Data Models. J. Econom.; 1998; 87, pp. 115-143. [DOI: https://dx.doi.org/10.1016/S0304-4076(98)00009-8]

51. Daud, S.N.M. Financial Inclusion, Economic Growth and the Role of Digital Technology. Financ. Res. Lett.; 2023; 53, 103602. [DOI: https://dx.doi.org/10.1016/j.frl.2022.103602]

52. Hansen, B.E. Threshold Effects in Non-Dynamic Panels: Estimation, Testing, and Inference. J. Econom.; 1999; 93, pp. 345-368. [DOI: https://dx.doi.org/10.1016/S0304-4076(99)00025-1]

53. Wang, Q. Fixed-Effect Panel Threshold Model Using Stata. Stata J.; 2015; 15, pp. 121-134. [DOI: https://dx.doi.org/10.1177/1536867X1501500108]

54. Luo, Y.; Lu, Z.; Wu, C.; Mensah, C.N. Environmental Regulation Effect on Green Total Factor Productivity: Mediating Role of Foreign Direct Investment Quantity and Quality. Int. J. Environ. Res. Public Health; 2023; 20, 3150. [DOI: https://dx.doi.org/10.3390/ijerph20043150]

55. Sajjad, F.; Zakaria, M. Credit Rating as a Mechanism for Capital Structure Optimization: Empirical Evidence from Panel Data Analysis. Int. J. Financ. Stud.; 2018; 6, 13. [DOI: https://dx.doi.org/10.3390/ijfs6010013]

56. Shao, M.; Fan, J.; Huang, Z.; Chen, M. The Impact of Information and Communication Technologies (ICTs) on Health Outcomes: A Mediating Effect Analysis Based on Cross-National Panel Data. J. Environ. Public Health; 2022; 2022, 2225723. [DOI: https://dx.doi.org/10.1155/2022/2225723]

57. Baron, R.M.; Kenny, D.A. The Moderator–Mediator Variable Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical Considerations. J. Pers. Soc. Psychol.; 1986; 51, pp. 1173-1182. [DOI: https://dx.doi.org/10.1037/0022-3514.51.6.1173]

58. MacKinnon, D.P.; Lockwood, C.M.; Hoffman, J.M.; West, S.G.; Sheets, V. A Comparison of Methods to Test Mediation and Other Intervening Variable Effects. Psychol. Methods; 2002; 7, pp. 83-104. [DOI: https://dx.doi.org/10.1037/1082-989X.7.1.83] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/11928892]

59. MaxDEA. MaxDEA Official Website. 2023; Available online: https://maxdea.com/Index_En.htm (accessed on 2 June 2025).

60. Smith, K.R. Allocating Responsibility for Global Warming: The Natural Debt Index. Ambio; 1991; 20, pp. 95-96. Available online: https://www.cabidigitallibrary.org/doi/full/10.5555/19911962207 (accessed on 5 June 2025).

61. You, J.; Xiao, H. Can FDI Facilitate Green Total Factor Productivity in China? Evidence from Regional Diversity. Environ. Sci. Pollut. Res. Int.; 2022; 29, pp. 49309-49321. [DOI: https://dx.doi.org/10.1007/s11356-021-18059-0]

62. Ishikawa, A. Why Does Production Function Take the Cobb–Douglas Form?. Statistical Properties in Firms’ Large-Scale Data; Springer: Singapore, 2021; pp. 113-135.

63. Li, X.; Hui, E.C.M.; Lang, W.; Zheng, S.; Qin, X. Transition from Factor-Driven to Innovation-Driven Urbanization in China: A Study of Manufacturing Industry Automation in Dongguan City. China Econ. Rev.; 2020; 59, 101382. [DOI: https://dx.doi.org/10.1016/j.chieco.2019.101382]

64. Wu, H.; Hao, Y.; Ren, S. How Do Environmental Regulation and Environmental Decentralization Affect Green Total Factor Energy Efficiency: Evidence from China. Energy Econ.; 2020; 91, 104880. [DOI: https://dx.doi.org/10.1016/j.eneco.2020.104880]

65. Yang, W.; Wang, D. Can Industrial Robot Applications Help Cross the Middle-Income Trap? Empirical Evidence Based on Cross-Country Panel Data. Technol. Forecast. Soc. Change; 2023; 192, 122583. [DOI: https://dx.doi.org/10.1016/j.techfore.2023.122583]

66. Wang, M.; Pang, S.; Hmani, I.; Li, C.; He, Z. Towards Sustainable Development: How Does Technological Innovation Drive the Increase in Green Total Factor Productivity?. Sustain. Dev.; 2021; 29, pp. 217-227. [DOI: https://dx.doi.org/10.1002/sd.2142]

67. Zhang, Q.; Zhang, F.; Mai, Q. Robot Adoption and Green Productivity: Curse or Boon?. Sustain. Prod. Consum.; 2022; 34, pp. 1-11. [DOI: https://dx.doi.org/10.1016/j.spc.2022.08.025]

68. Yang, G.; Hou, Y. The Use of Industrial Robots, Technology Upgrade and Economic Growth. China Ind. Econ.; 2020; 10, pp. 138-156. [DOI: https://dx.doi.org/10.19581/j.cnki.ciejournal.2020.10.004]

69. Zhang, P.; Qin, Y.; Liang, H.; Zhou, L. Robotization and Labour Demand in the Post-Pandemic Era: Microeconomic Evidence from China. Technol. Forecast. Soc. Change; 2023; 192, 122523. [DOI: https://dx.doi.org/10.1016/j.techfore.2023.122523]

70. Wang, H.; Wang, Y.; Zeng, G.; Qian, Z.; Lu, S. The Differential Impact of the Digital Economy on Urban Energy Efficiency in China: The Mediating Mechanism of FDI. Environ. Dev. Sustain.; 2024; 26, pp. 31323-31350. [DOI: https://dx.doi.org/10.1007/s10668-024-04475-y]

71. Zhao, M.; Liu, F.; Sun, W.; Tao, X. The Relationship between Environmental Regulation and Green Total Factor Productivity in China: An Empirical Study Based on the Panel Data of 177 Cities. Int. J. Environ. Res. Public Health; 2020; 17, 5287. [DOI: https://dx.doi.org/10.3390/ijerph17155287]

72. Pugliese, S. Toward a Multilevel System of Investment Control Oriented to Crisis Management: Italian Golden Power in the Framework of the EU FDI Screening Mechanism. National Security and Investment Controls; Pohl, J.H.; Papadopoulos, T.; Wiesenthal, J. Springer Studies in Law & Geoeconomics, Volume 3; Springer Nature: Cham, Switzerland, 2024; pp. 231-261.

73. Tang, D.; Shan, Z.; He, J.; Zhao, Z. How Do Environmental Regulations and Outward Foreign Direct Investment Impact the Green Total Factor Productivity in China? A Mediating Effect Test Based on Provincial Panel Data. Int. J. Environ. Res. Public Health; 2022; 19, 15717. [DOI: https://dx.doi.org/10.3390/ijerph192315717]

74. Chen, C.; Lan, Q.; Gao, M.; Sun, Y. Green Total Factor Productivity Growth and Its Determinants in China’s Industrial Economy. Sustainability; 2018; 10, 1052. [DOI: https://dx.doi.org/10.3390/su10041052]

75. Wang, Y.; Xie, L.; Zhang, Y.; Wang, C.; Yu, K. Does FDI Promote or Inhibit the High-Quality Development of Agriculture in China? An Agricultural GTFP Perspective. Sustainability; 2019; 11, 4620. [DOI: https://dx.doi.org/10.3390/su11174620]

76. Cheng, P.; Wang, X.; Choi, B.; Huan, X. Green Finance, International Technology Spillover and Green Technology Innovation: A New Perspective of Regional Innovation Capability. Sustainability; 2023; 15, 1112. [DOI: https://dx.doi.org/10.3390/su15021112]

77. Chen, S.; Golley, J. ‘Green’ Productivity Growth in China’s Industrial Economy. Energy Econ.; 2014; 44, pp. 89-98. [DOI: https://dx.doi.org/10.1016/j.eneco.2014.04.002]

78. Lee, K.-H.; Min, B. Green R&D for Eco-Innovation and Its Impact on Carbon Emissions and Firm Performance. J. Clean. Prod.; 2015; 108, pp. 534-542. [DOI: https://dx.doi.org/10.1016/j.jclepro.2015.05.114]

79. Akisik, O.; Gal, G.; Mangaliso, M.P. IFRS, FDI, Economic Growth and Human Development: The Experience of Anglophone and Francophone African Countries. Emerg. Mark. Rev.; 2020; 45, 100725. [DOI: https://dx.doi.org/10.1016/j.ememar.2020.100725]

80. Fatorachian, H.; Kazemi, H. A Critical Investigation of Industry 4.0 in Manufacturing: Theoretical Operationalisation Framework. Prod. Plan. Control; 2018; 29, pp. 633-644. [DOI: https://dx.doi.org/10.1080/09537287.2018.1424960]

© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.

Industrial Robots and Green Productivity: Evidence from Global Panel Data on High-Quality Economic Development

Content area

Abstract

Full text

1. Introduction

2. Conventional Facts, Theoretical Framework, and Hypotheses

2.1. Total Factor Productivity and Its Green Transformation

2.2. Classification of Countries Based on Economic Development

2.3. Heterogeneous Impact of IRAs on HQED in Developing and Developed Countries

2.4. Theoretical Examination and Hypothesis Development for IRAs in the Context of HQED

2.5. Research Gaps and Theoretical Contributions

3. Methodology

3.1. Model Selection

3.1.1. GTFP Measurement

3.1.2. Fixed Effects and the System Generalized Method of Moments Model

3.1.3. Panel Threshold Model

3.1.4. Mediation Effect Model

3.2. Descriptive Statistics and Data Source

3.2.1. Descriptive Statistics

3.2.2. Data Sources

4. Empirical Analysis

4.1. Heterogeneity Analysis Across Groups

4.2. Exploring the Threshold Effect of IRAs on ERs in Relation to GTFP Enhancement

4.3. Robustness Test

4.3.1. Effect of IRAs on GTFP

4.3.2. Threshold Model Robustness

4.4. Analysis of Mechanisms

4.4.1. Mechanism of FDI

4.4.2. Mechanism of TI

4.4.3. R&D Mechanism

5. Conclusions and Implications

5.1. Conclusions

5.2. Implications

5.2.1. In Developing Countries

5.2.2. In Developed Countries

5.2.3. For Developing Nations

5.2.4. For Developed Nations