Biased innovation and network evolution: Digital driver for green innovation of manufacturing in China

Liu, Yang; Cheng, Jing; Dai, Jingjing Article Biased innovation and network evolution: Digital driver for green innovation of manufacturing in China Journal of Applied Economics Provided in Cooperation with: University of CEMA, Buenos Aires Suggested Citation: Liu, Yang; Cheng, Jing; Dai, Jingjing (2024) : Biased innovation and network evolution: Digital driver for green innovation of manufacturing in China, Journal of Applied Economics, ISSN 1667-6726, Taylor & Francis, Abingdon, Vol. 27, Iss. 1, pp. 1-31, https://doi.org/10.1080/15140326.2024.2308951 This Version is available at: https://hdl.handle.net/10419/314255 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/4.0/ Journal of Applied Economics ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/recs20 Biased innovation and network evolution: digital driver for green innovation of manufacturing in China Yang Liu, Jing Cheng & Jingjing Dai To cite this article: Yang Liu, Jing Cheng & Jingjing Dai (2024) Biased innovation and network evolution: digital driver for green innovation of manufacturing in China, Journal of Applied Economics, 27:1, 2308951, DOI: 10.1080/15140326.2024.2308951 To link to this article: https://doi.org/10.1080/15140326.2024.2308951 © 2024 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. Published online: 29 Jan 2024. Submit your article to this journal Article views: 731 View related articles View Crossmark data Citing articles: 2 View citing articles Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=recs20 RESEARCH ARTICLE Biased innovation and network evolution: digital driver for green innovation of manufacturing in China Yang Liu a , Jing Cheng b and Jingjing Dai a a School of Economics and Management, Kunming university, Kunming, China; b School of Construction Engineering, Yunnan Agricultural University, Kunming, China ABSTRACT The study aims to explore the spatial association network characteristics of biased green innovation in the manufacturing sector and its core drivers. This study constructs a Malmquist-Luenberger decomposition index model to identify the input and output biases of green technological innovation (GIIM and GIOM) in the manufacturing industry. This study uses a modified gravity model and social network analysis method to conduct a robust assessment of GIIM spatial association network of 30 provinces in China from 2012 to 2021. The results show: (1) The GIIM association network structure is stable and has good accessibility, with close connections between provinces and blocks, and significant spillover effects between provinces. (2) The regional network shows a “coreperiphery” spatial variation, with the core area expanding and the peripheral area shrinking. (3) The digital transformation characteristics of the network components and the intensity of environmental regulation have a significant impact on GIIM. ARTICLE HISTORY Received 3 September 2023 Accepted 10 January 2024 KEYWORDS social network analysis; spatial and evolutionary analysis; biased green innovation; digital transformation 1. Introduction The manufacturing sector has been instrumental to China’s economic expansion, contributing nearly 30% to the nation’s GDP in 2021, a notable proportion compared to other major global economies, as reported by the National Bureau of Statistics. Despite this success, the sector casts a long shadow of environmental concerns. It is a major consumer of energy and a major emitter of carbon, accounting for over 30% of China’s total energy consumption and approximately 35% of its total carbon emissions as of 2020. These statistics underscore the urgent need for a sustainable shift (W. Chen et al., 2017; Qin et al., 2021). In alignment with the “Made in China 2025” strategy, green development has been identified as a foundational principle for the manufacturing industry, advocating for the widespread implementation of green manufacturing practices to facilitate the industry’s ecological transformation (Wübbeke et al., 2016). At the heart of this transformative agenda is green technological innovation (Yuan & Xiang, 2018), which, through the adoption of eco-friendly production technologies, has CONTACT Jing Cheng [email protected] School of Construction Engineering, Yunnan Agricultural University, No.452, Feng Yuan Road, Pan Long District, Kunming, China JOURNAL OF APPLIED ECONOMICS 2024, VOL. 27, NO. 1, 2308951 https://doi.org/10.1080/15140326.2024.2308951 © 2024 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License (http:// creativecommons.org/licenses/by-nc/4.0/), which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. The terms on which this article has been published allow the posting of the Accepted Manuscript in a repository by the author(s) or with their consent. the potential to significantly curb energy consumption and pollution. Such innovations not only promise enhanced energy efficiency and cost reductions but also bolster the long-term competitiveness of the manufacturing sector (Herring & Roy, 2007). In the context of a developing nation like China, the pivot towards green innovation is key, promising environmental improvement while also carving out novel market and growth avenues (Baloch et al., 2021). The transition to green technological innovation necessitates a departure from traditional resource-intensive methods to those that prioritize technology and intelligent inputs. Innovation with an environmental bias is essential in steering this change. He and Wang’s (2015) research corroborates that innovation with such a bias can improve corporate environmental performance and foster sustainable development (He & Wang, 2015). In today’s climate, directing innovation resources toward the development and implementation of green technologies is imperative for the green economic transformation of China. Moreover, the innovation of green technology in the manufacturing sector is embedded within a dynamic spatial network, influenced by an interplay of policy, market, technology, and environmental factors (Malecki & Edward, 2008). A firm’s innovative endeavours are significantly shaped by its regional innovation ecosystem (Audretsch et al., 2022; Radziwon et al., 2022). With advancements in information and digital technologies, traditional geographical limitations are becoming less pronounced, leading to a complex network of cross-regional spatial interactions within China. This study hypothesizes that green innovation in China’s manufacturing sector is not only contingent upon the direction of green technology innovation bias but is also characterized by a pronounced spatial dependency. To sum up, this research endeavours to dissect the inclination towards green technological innovation within the manufacturing realm, elucidating its spatial network dynamics and identifying the catalysts behind it. By clarifying these aspects, we seek to provide a clearer picture of the opportunities and challenges confronting China’s manufacturing sector on its path to a green future. The structure of the study is as follows: Section 2 provides a comprehensive literature review, contextualizing our study within the existing body of work on bias in technological innovation and the networks within innovation spaces. Section 3 outlines the research methodology and delineates the data sources utilized. Section 4 presents the core analytical findings along with pertinent discussion, which focuses on the spatial correlation of input-biased green innovation in the manufacturing sector. Section 5 conducts an empirical analysis of drivers. Finally, Section 6 summarizes the primary conclusions, discussions and offers corresponding policy recommendations. 2. Literature review 2.1. Research related to biased technological innovation David and van de Klundert (1965) study was a seminal work in gauging the trajectory of technological progress, using the CES production function to evaluate the factor bias of technology (David & van de Klundert, 1965). Later, Klump in 2012 employed the standardized supply-side system approach to quantify the orientation of technological innovations (Klump et al., 2012). However, as the complexity of input factors grew, the 2Y. LIU ET AL. fixed elasticity substitution of the CES function fell out of favor, with the Translog production function becoming the preferred tool for scholars to measure the directionality of technological progress. For example, Karanfil and Yeddir-Tamsamani (2010) utilized the Translog production function to assess energy-focused technological innovations in France, pinpointing energy prices as the key determinant of directional bias (Karanfil & Yeddir-Tamsamani, 2010). Subsequent research, starting with Fare et al., began incorporating nonparametric methods to measure directional bias (Färe et al., 1997). This involved analyzing shifts in the production frontier over time that are nonproportional, leading to alterations in the marginal output ratios of various factors. A common technique is the DEA-Malmquist index method, which evaluates technological progress from both input and output dimensions, examining the influence of its directional components on total factor productivity (P.-C. Chen & Yu, 2014; Färe et al., 1997; Peng et al., 2019; Weber & Domazlicky, 1999). The Malmquist index method mitigates the subjective bias that can arise from production function assumptions and reveals the impact of the orientation of technological innovation on different inputs. Yet, when it comes to green technological innovation, it’s imperative to acknowledge the variety of input factors and the potential for environmentally detrimental outputs. Chinese researchers, like Yang et al. (2019), have innovated by integrating the SBM directional distance function with the Malmquist-Luenberger index, formulating a novel methodology to gauge green technological progress. Building upon this foundation, our study factors in non-desired outputs and undertakes a comprehensive decomposition of the Malmquist-Luenberger index. This allows us to measure the intergroup bias of green technological innovation within the context of resource and environmental impacts, examining both the inputs and the outputs. This multidimensional approach provides a more nuanced understanding of the green innovation trajectory, reflecting the intricate balance between economic progress and environmental stewardship. 2.2. Research related to innovation networks The field of spatial network analysis emerged from Castells’ Space of Flows theory, which posits that society is a spatial form constituted by the flows of various elements (Castells & Cardoso, 2006). Building on this, the globalization and world cities research network has deepened the study of spatial networks by linking elements such as infrastructure and corporate organization with spatial data (Derudder & Taylor, 2021). Scholars like Rogers (2004) and Feldman (2016) have further enriched this domain by introducing innovation into spatial network studies, uncovering the significant impact of network structures on the diffusion of innovation. For instance, Huggins et al. (2023) examined the relationship between the regional concentration and dispersion of innovative agents within specific industries and their capacity for innovation. Research on the spatial effects of green innovation frequently employs spatial econometric methods, utilizing tools like the Moran’s I (Gai et al., 2022), kernel density estimation (P. Zhao et al., 2023), and Theil index (N. Zhao et al., 2021) to investigate the spatiotemporal evolution and differentiation of green innovation (Liang et al., 2022; Xuhui & Yitao, 2023; Zhou et al., 2021). JOURNAL OF APPLIED ECONOMICS 3 Recent studies on green innovation predominantly rely on “attribute data”, which, as Z. Chen et al. (2022) note, inadequately capture the interactive mechanisms between regions. Analysing spatial association networks from a relational viewpoint, as advocated by CHONG and QIN (2017), yields more insightful findings than merely examining “attribute data” through traditional spatial econometric methods. This relational perspective reveals a consensus in the literature: regions with higher innovation efficiency often occupy central positions in these networks. These regions leverage locational advantages to access extensive innovative resources, enhancing their innovative efficiency (Feng et al., 2022; Guan et al., 2015). In contrast, regions with lower innovation efficiency benefit less from network effects, leading to uneven regional innovation development (Min et al., 2020). In the context of China’s sustainable development goals, the equitable regional distribution of green innovation efficiency is paramount. However, research by SUN et al. (2022) indicates a persistent core-periphery structure in green technological innovation, pointing to a significant relationship between directed investment in green technology innovation elements and innovation outcomes. This relationship, coupled with the increasing degree of interregional connectivity, suggests that a region’s green innovation propensity is increasingly influenced by its neighbours’ activities. Despite the recognition of spatial correlation in green innovation, existing research has shortcomings that necessitate comprehensive resolution. Traditional spatial exploratory analyses and econometric models primarily focus on the heterogeneity in spatial distribution due to geographical proximity. These approaches, however, overlook the analysis of internal biases and regional holistic perspectives, thereby constraining the understanding of the characteristics and generative mechanisms of spatial correlation networks. The predominant focus on attribute data, with scant exploration of relational data in complex spatial networks, leads to inadequate differentiation of the roles, functions, and mechanisms of different regions within the green innovation bias network. Therefore, in the face of uneven green innovation efficiency and strengthened spatial correlations, employing social network analysis offers a more comprehensive exploration of interregional relationships. This method not only unravels the mechanisms of spatial relationships but also supports more rational input allocation, providing insights for promoting balanced development in green innovation. This study examines 30 provincial-level administrative regions, employing a Malmquist-Luenberger decomposition index model to assess the bias in green innovation in manufacturing (GIBM) from 2012 to 2021. GIBM refers to a tendency, inclination, or preference that is embedded within eco-innovation. Additionally, it adopts a modified gravity model to construct an input biases of green technological innovation (GIIM) spatial matrix, a spatial matrix designed to measure the regional disparities in green technological inputs biases. This research analyses the characteristics of the spatial network structure of GIIM using social 4Y. LIU ET AL. network methods. Furthermore, QAP regression is utilized to investigate the drivers of GIIM spatial correlation networks, thereby elucidating the formation mechanism of these spatial correlation networks (Krackardt, 1987). 3. Materials and methods 3.1. Measurement of the GIBM 3.1.1. Construction of the indicator system Drawing upon relevant literature (Egilmez et al., 2013; Sims et al., 1974; Zheng et al., 2021), this study defines the key components in the context of manufacturing production as follows: Input factors are identified as four key elements – capital, land, labour, and energy. The desired output is conceptualized primarily as economic output indicators. However, this study also includes technological innovation in the desired output, reflecting a more realistic approach to contemporary demands. The non-desired output typically encompasses waste emissions (Kaneko & Managi, 2004), including a range of pollutants such as Chemical Oxygen Demand (COD), ammonia emission, CO2 emissions, sulfur dioxide, nitrogen oxide, wastewater, and solid waste (Liu et al., 2022; Yang et al., 2019; B. Zhang et al., 2022). Detailed indicators are outlined in Table 1. 3.1.2. Measurement method We assume there are N decision-making units (DMUs) in the manufacturing, each utilizing W types of inputs, producing Q types of expected outputs, and O types of unexpected outputs in each period. For the nth n¼1...Nð Þ DMU in period t t ¼1;2...;Tð Þ, the set of input factors is denoted as xt n¼xt 1n;...;xt Wn  �, the set of expected output factors as yt n¼yt 1n;...;yt Qn � �, and the set of unexpected output factors as bt n¼bt 1n;...;bt On  �. If the Directional Distance Function (DDF) satisfies: Dt oxt;yt;bt;yt;bt  �¼max β Table 1. GIBM evaluation index system. Guideline Layer Indicator Layer Proxy Indicators Input factors Capital input Investment in fixed assets in manufacturing (billion yuan) Labor input Employment in manufacturing (million people) Land input Industrial land area (km 2 ) Energy input Comprehensive energy consumption per unit of industrial added value (t standard coal/million yuan) Expected Output Economic output Industrial added value as a share of GDP (%) Technology output Number of green patents of manufacturing companies Unexpected Output COD emissions COD emission per unit of industrial added value (t/million yuan) (%) Ammonia emission Ammonia nitrogen emission per unit of industrial added value (t/100 million yuan) (%) Sulfur dioxide emission SO 2 emissions per unit of industrial added value/(t/billion yuan) (%) Nitrogen oxide emissions NO x emission per unit of industrial added value/(t/billion yuan) (%) Wastewater emissions Wastewater emission per unit of industrial added value/(t/billion yuan) (%) Solid waste production Solid waste generation per unit of industrial added value/(t/yuan) (%) JOURNAL OF APPLIED ECONOMICS 5 P N n¼1 zt nxt wn �xt wn;w¼1;2;���;W P N n¼1 zt nyt wn �1þβð Þyt qn;q¼1;2;���;Q P N n¼1 zt nbt wn ¼1βð Þbt on;o¼1;2;���;O zt n�0;n¼1;2;. . . ;N 8 > > > > > > > > > < > > > > > > > > > : (1) Where β gauges the distance of the evaluated DMU to the production frontier in the direction of yt;bt  �:By integrating the constraint condition of Equation (1) with the Malmquist-Luenberger (ML) index method, the ML index can be decomposed into efficiency change (ΔT) and technological change (ΔTE). Further, the ΔTE index is decomposed to obtain the input-biased of green technological innovation in manufacturing (GIIM) and the output bias of green technological innovation (GIOM) (Färe et al., 1997). The specific calculation method is as follows: Mt 0xt;yt;xtþ1;ytþ1  �¼Dt 0xtþ1;ytþ1  �=Dt 0xt;yt ð Þ;t¼1;...;T1 ¼Δ T xtþ1;ytþ1  ��Δ TE xt;yt;xtþ1;ytþ1  � ¼Dt 0xtþ1;ytþ1 ð Þ Dtþ1 0xtþ1;ytþ1 ð Þ � ��Dtþ1 0xtþ1;ytþ1 ð Þ Dt 0xt;yt ð Þ � �: GIIM xt;yt;xtþ1  �¼Dtþ1 iyt;xtþ1 ð Þ Dt iyt;xtþ1 ð Þ =Dtþ1 iyt;xt ð Þ Dt iyt;xt ð Þ � � (2) GIOM yt;xtþ1;ytþ1  �¼Dt 0xtþ1;ytþ1 ð Þ Dtþ1 0xtþ1;ytþ1 ð Þ=Dt 0xtþ1;yt ð Þ Dtþ1 0xtþ1;yt ð Þ � � Factor input bias is measured based on the change in the marginal rate of substitution of factors (Weber & Domazlicky, 1999). This study examines the energy substitution bias (ESB) index of green technology innovation in manufacturing to determine the network node attributes. The ESB is expressed as follow: ESBC;E¼Ctþ1 Etþ1=Ct Et1 � ��GIIM 1ð Þ (3) where C refers to capital input and E refers to energy input. When there is technological innovation from period t to tþ1, Ctþ1 Etþ1=Ct Et is the ratio of the marginal substitution rates of factors C and E from stage t to tþ1, reflecting the factor changes in innovation. When GIIM >1, Ctþ1 Etþ1=Ct Et>1, that is ESBI;J>0, it indicates the energy-saving green technology innovation. 3.2. Network construction and network analysis methods 3.2.1. Network construction The identification of spatial linkages is crucial for analyzing the spatial linkage network of GIBM in China using social network methodologies (Borgatti & Foster, 2003; Cassi et al., 6Y. LIU ET AL. 2012). Given that spatial linkages often diminish with increasing geographical distance – a phenomenon known as “distance decay” highlighted by Basile et al. (2012)—the closer the geographical proximity, the stronger the linkages tend to be. Therefore, this study proposes the use of a modified gravity model to construct the spatial association network. This approach integrates considerations of green manufacturing with economic and geographical distances, thereby offering a more nuanced understanding of the spatial association characteristics. Additionally, the economic distance between regions is a significant factor influencing these linkages. Incorporating population and economic scale into the gravity model (Кузнецов et al., 2014), allows for a more accurate depiction of the spatial evolution trends. In this research, we adapt the gravity model to determine the spatial correlation network of green manufacturing among different regions. The specific calculation formula employed is outlined as follows: Fij ¼Kij ffiffiffiffiffiffiffiffiffiffiffiffiffiffi PiGiMi 3 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi PjGjMj 3 p D2 ij ;Kij ¼Mi MiþMj ;Dij ¼dij gigj (4) In Equation (4): Fij denotes the spatial correlation strength (gravitational value) of GIBM between provinces i and j; Mi and Mj denote the green manufacturing development indexes of provinces i and j, respectively; Kij denotes the contribution rate of province i to Fij; dij is the geographic distance between provinces i and j; Gi and Gj denote the level of economic development of provinces i and j, as measured by the total GDP; gi and gj are the per capita GDP of the two regions, respectively. Building upon this framework, this study acknowledges the presence of a threshold value in the strength of spatial relationships. Consequently, this study proposes using the average value of each row in the association strength matrix derived from the gravity model as this threshold value. Following this, a 0–1 matrix is constructed, determining the presence or absence of association relationships. This process culminates in the formation of a directed binary spatial association matrix. 3.2.2. Social network analysis methods This study delves into the spatial association network of GIBM, employing social network analysis as a framework (Fritze et al., 2018). In examining the overall structure of the network, we focus on several key characteristics. Network Density (ND) is used to gauge the complexity of the relationships within the green manufacturing association network. It reflects the intricacy of connections among various nodes. The number of nodes and the quantity of relationships within the network indicate the network’s relevance (NR), which is a measure of the GIBM network structure’s stability. Network Efficiency (NE) is employed to quantify the characteristics of GIBM association channels. Furthermore, Network hierarchy (NH) is analysed to understand the degree of asymmetric accessibility within the network. This metric is crucial in evaluating the hierarchical structure of access to resources and information. The presence of small-world characteristics within the network is scrutinized to assess the efficiency and accessibility of resource dissemination across the network. The calculations for ND, NR, NE and NH is as presented in Eq. 5–8: ND ¼N=TN �TN 1ð Þ½ � (5) JOURNAL OF APPLIED ECONOMICS 7 The indicators in Figure 7 collectively demonstrate not only the quantitative growth of the network but also its qualitative development. The expansion in network density and relationships signifies a maturing landscape of GIIM in China, reflecting a nationwide shift towards more collaborative and interconnected approaches in this field. This trend is a positive indicator of the progress in the application and integration of green technologies across the provinces, marking a significant step towards achieving sustainable development goals. The paper further calculates key metrics such as the degree of correlation, network hierarchy, and network efficiency within the GIIM spatial correlation network. As detailed in Table 3, the network correlation degree was consistently measured at 1 throughout the study period. This consistency suggests a relatively stable network structure, despite minor fluctuations in the number of network relationships. Such stability indicates good network accessibility among the provinces, characterized by strong spatial correlations and spillover effects, implying that most provinces have established stable cooperative relationships in the realm of GIIM. Additionally, the network hierarchy degree, measured at 0, reveals that the GIIM spatial correlation network operates on a relatively flat hierarchical scale at the interprovincial level. This absence of a strict hierarchical order and connectivity barriers underlines a significant synergistic effect among the provinces, facilitating collaborative and egalitarian interactions in green innovation. Figure 7. Network density and association from 2012 to 2021. Table 3. Connectedness, hierarchy and efficiency. year 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 Connectedness 1 1 1 1 1 1 1 1 1 1 Hierarchy 0 0 0 0 0 0 0 0 0 0 Efficiency 0.5394 0.4901 0.4557 0.4286 0.4187 0.4212 0.3768 0.3842 0.3227 0.3128 14 Y. LIU ET AL. The overall trend of declining network efficiency, with distinct phase characteristics, indicates that the channels for GIIM spatial spillover have expanded during the study period. This expansion has ostensibly enhanced the overall stability of the network, suggesting a maturing and more interconnected GIIM landscape. 4.4. Networked small-world features In this study, we constructed a random network that matches the scale and density of real-world correlation networks annually. We characterized the connectivity between network nodes using metrics such as network clustering coefficient, average distance, and network diameter. These analyses are illustrated in Figure 8. From 2012 to 2021, the network’s clustering coefficient remained relatively stable, fluctuating between 0.6 and 0.8. This indicates a strong degree of connectivity among nodes and their adjacent nodes. The spatial correlation network of the GIIM maintained a balanced state of connectivity, aligning with the trends observed in the GIIM spatial correlation networks depicted in Figures 4–6. The average path length varied between 1.544 and 1.321, showing a yearly decreasing trend. This decrease suggests that provinces have been leveraging the network structure for more effective collaboration in GIIM, thereby reducing network redundancy and enhancing overall efficiency. Remarkably, the network diameter consistently remained at 2, significantly lower than the total number of 30 members in the overall spatial network. This consistency highlights the network’s efficient structure, where the distance between any two nodes is relatively small. Overall, the GIIM spatial correlation network exhibits characteristics of a “smallworld” phenomenon, indicative of excellent network connectivity. This phenomenon allows for the rapid dissemination of a province’s level of GIIM to others, thereby Figure 8. Evolution of small-world characteristics of spatially linked networks. JOURNAL OF APPLIED ECONOMICS 15 Figure 9. Centrality analysis in 2012. Figure 10. Centrality analysis in 2016. 16 Y. LIU ET AL. improving the spatial correlation network status of provinces with lower GIIM levels. These findings underscore the vital role of cross-provincial and cross-regional collaboration in fostering the development and dissemination of GIIM, highlighting the interconnected nature of green innovation efforts across China. 4.5. Ego-network characteristics and dynamic evolution trend 4.5.1. Centrality analysis To elucidate the positions and functions of each province within the GIIM spatial association network, this study quantitatively assesses three pivotal network metrics : degree centrality, betweenness centrality, and closeness centrality. The years 2012, 2016, and 2021 were specifically chosen for this dynamic analysis, with Figures 9–11 illustrating the distribution of individual network characteristics related to spatial associations across the provinces. In terms of degree centrality, regions such as Beijing, Shanghai, Jiangsu, and Zhejiang consistently emerged as central players in the GIIM network throughout the research period. The stable yet notable increase in both in-degree and out-degree for various provinces, particularly in 2021, suggests an intensifying spillover effect. Notably, Gansu’s higher in-degree compared to its out-degree each year implies it benefits more from the network than it contributes, while Beijing and Tianjin, with high in-degrees, demonstrate a strong capacity to absorb resources and promote local GIIM development through a “siphoning effect”. Figure 11. Centrality analysis in 2021. JOURNAL OF APPLIED ECONOMICS 17 The average betweenness centrality value decreased over the years, indicating a gradual reduction in bipolarization within the GIIM network among provinces. This trend is beneficial for enhancing cooperative efforts and communication in GIIM development. Jiangsu, Beijing, and Shanghai, consistently ranking high in betweenness centrality, play crucial roles due to their strategic positions in inland communication channels. Closeness centrality rankings closely align with degree centrality, suggesting that provinces with significant positions and close ties to others also influence the GIIM of neighboring provinces. The noticeable upward trend in closeness centrality, with more provinces exceeding the average value each year, indicates an enhanced ability to connect through shorter paths and access innovation resources effectively. By combining these three centrality indices to examine green innovation in the manufacturing sectors of various provinces, an increasingly strong investment bias is evident. Regions with advanced manufacturing, such as Shanghai, Beijing, and Jiangsu, continue to exert strong control over GIIM-related innovation resources. However, the intensification of spatial network connections also brings geographically more remote areas into prominence within these networks. 4.5.2. Structural hole analysis Applying the theory of structural holes to the GIIM network, this study measures the positions of structural holes within the network, using two indicators: Effective Size and Constraint. “Effective Size” gauges the extent of a province’s influence on the overall GIIM network, while “Constraint” assesses the likelihood of a province occupying a structural hole position (Y. Chen, 2015). In a representation (Figure 12), the study examines advantageous and disadvantageous nodes in the context of green innovation in 2021. It reveals that regions like Beijing, Zhejiang, and Fujian, with large Effective Sizes and low Constraints, are Figure 12. Structural hole indicator of China’s GIIM in 2021. 18 Y. LIU ET AL. positioned advantageously in structural holes. This positioning implies a greater ability to access diverse information and exert control within the network. Conversely, regions such as Xinjiang, Hainan, and Jilin, characterized by small Effective Sizes and high Constraints, demonstrate weaker external resource capabilities and a high dependency on provinces occupying advantageous structural hole positions. These findings indicate a disparity in the capacity to leverage network benefits for green innovation. The study suggests that regions in advantageous node positions should capitalize on their network benefits in green manufacturing innovation. By enhancing information exchange with regions in disadvantaged node positions, they can facilitate a more balanced and inclusive regional development. This strategy is crucial for ensuring that all provinces, regardless of their current network position, can contribute to and benefit from advancements in green technology innovation. 4.6. Evolution of core-periphery structure Utilizing core-periphery analysis, it is evident that the spatial correlation network of China’s GIIM is characterized by an expanding core and a diminishing periphery, as illustrated in Figure 13. From 2012 to 2021, the number of provinces in the core region steadily increased, from 17 in 2012 to 19 by 2021. This increase reflects a significant trend of gradual extension from the south-eastern coastal provinces to more peripheral areas, marking a shift from a single-core agglomeration to a more diverse, multi-core expansion. The evolving distribution of the core region signifies not only the geographic spread of green innovation but also the increasing integration and collaboration among provinces. This expansion demonstrates the growing influence and reach of green innovation initiatives, particularly from the developed provinces along the eastern coast. Concurrently, the diminishing periphery illustrates how previously peripheral provinces are becoming increasingly integrated into the network. This integration is driven by enhanced economic development and inter-provincial connections, leading to tighter ties within the density matrix. As a result, peripheral provinces are gradually becoming active participants in the GIIM network, contributing to and benefiting from green innovation initiatives. Figure 13. Core-edge structure of GIIM spatial associations, 2012, 2016 and 2021. JOURNAL OF APPLIED ECONOMICS 19 The developed provinces along the eastern coast play a pivotal role in this process. Their expansion in diffusion and radiation scope significantly promotes the improvement of the GIIM level in surrounding provinces. This trend is indicative of a broader shift towards a more inclusive and interconnected approach to GIIM, where provinces show increasing initiative and capability in contributing to the network structure. Overall, while the spatial correlation network of China’s GIIM continues to exhibit a core-periphery structure, the dynamics within this structure are rapidly evolving. The increasing number of core areas and the integration of peripheral regions underscore a national trend towards more integrated and collaborative green innovation efforts. 4.7. Blocks analysis The analysis of China’s GIIM, considering both overall and individual characteristics, reveals significant spatial differentiation. To further depict the inter-regional interactions and dynamic changes, this study selects data from 2012, 2016, and 2021. An iterative algorithm, which considers segmentation depth and concentration standards, categorizes Figure 14. Correlation among the four GIIM blocks in 2021. Table 4. Classification of GIBM space-related network blocks in 2012. Segment Segment Matrix provinces in each segment overflow relationships accepted relationships Desired internal relationship ratio Actual internal relationship ratioI II III IV I 2 12 24 20 2 56 55 3.45% 3.45% II 12 21 69 42 6 123 106 17.24% 4.65% III 23 68 40 17 12 108 109 37.93% 10.00% IV 20 26 16 15 10 62 79 31.03% 13.89% 20 Y. LIU ET AL. 30 regions into four major blocks. The results show R 2 values of 0.531 (2012), 0.557 (2016), and 0.519 (2021), indicating a generally good fit for each year. Figure 14 illustrates the interactive relationships between these four major blocks of the GIIM spatial correlation network in 2021. The block model analysis results, as listed in Tables 4–6, reveal that when spatial correlations are considered, inter-block relationships are significantly more prevalent than intra-block associations. This indicates notable spatial correlations and spillover effects between GIIM blocks. By comprehensively considering the number of receiving blocks, spillover relationship counts, internal relationship counts, and the proportion of internal relationships, the attributes of the blocks for 2012, 2016, and 2021 were categorized. A closer examination of their dynamic changes reveals: ●Block I: Weak internal spillover effects, significant spillover to others, and reception of spillover from other blocks, playing an effective “bridge” role in the spatial correlation network. Throughout the study period, it was consistently categorized as the “Broker” block. ●Block II: The number of spillover relationships exceeds those received, making it a “Net Spillover” block throughout the study period. This block not only meets its own development needs but also facilitates the development of other block members through element spillover. ●Blocks III and IV: These blocks experienced role reversals during the study period. In 2012 and 2016, Block III, with a roughly equal number of received and spillover relationships and a relatively high proportion of internal relationships, was categorized as a “Bidirectional Spillover” block, playing a “Bidirectional Guide” role. In 2021, it became a “Net Inflow” block, whereas Block IV transitioned from a “Net Inflow”block in 2012 and 2016 to a “Bidirectional Spillover” block in 2021. Not only did the attributes of the blocks change, but also the composition of provinces within each block shifted notably. For instance, the number of provinces in Block Table 5. Classification of GIBM space-related network blocks in 2016. Segment Segment Matrix provinces in each segment overflow relationships accepted relationships Desired internal relationship ratio Actual internal relationship ratioI II III IV I 7 24 39 18 3 81 81 6.90% 3.57% II 24 33 97 33 8 154 142 24.14% 4.94% III 39 97 57 5 13 141 139 41.38% 8.44% IV 18 21 3 12 6 42 56 17.24% 12.50% Table 6. Classification of GIBM space-related network blocks in 2021. Segment Segment Matrix provinces in each segment overflow relationships accepted relationships Desired internal relationship ratio Actual internal relationship ratioI II III IV I 30 29 54 60 6 143 144 17.24% 4.03% II 30 12 45 43 5 118 112 13.79% 4.07% III 54 41 21 37 9 132 140 27.59% 6.38% IV 60 42 41 22 10 143 140 31.03% 6.54% JOURNAL OF APPLIED ECONOMICS 21 I increased, with Beijing, Tianjin, Jiangsu, Zhejiang, Fujian, and Shanghai being part of the “Broker””block by 2021, indicating an increasing number of provinces playing a“bridge”- and“intermediary”role. The variations in the provincial composition of the other blocks across different years highlight the complexity and dynamism of the GIIM spatial correlation network. 4.8. Network structure effect analysis To reveal the structural characteristics of the GIIM spatial correlation network, this study empirically examines the impact of network structure on regional differences and levels of GIIM from two aspects: the overall network and individual network structures. From the perspective of the overall network structure, the variation coefficient of GIIM levels across provinces is used to measure the inter-provincial differences in GIIM levels. OLS regression is conducted using network density and network efficiency, with the dependent variables undergoing logarithmic transformation. Additionally, from the perspective of individual network structure, besides the degree centrality and betweenness centrality used in the individual characteristics analysis, eigenvector centrality is also introduced as an explanatory variable. Balanced panel data regression analysis is performed with GIIM as the dependent variable. Logarithmic transformation of the explanatory variables is carried out to avoid certain multicollinearity issues. 4.8.1. Regression results for the whole network structure Table 7 reveals that the regression coefficients for network density and network efficiency on the regional differences of GIIM are 2.420 and −2.357, respectively, and both are significant at the 5% level. This indicates that an increase in network density and a decrease in network efficiency significantly affect the regional disparities of GIIM, leading to a more balanced spatial distribution of GIIM levels. The potential reasons are as follows: Firstly, an increase in network density enhances the cohesion among provinces in terms of green innovation orientation, thus avoiding spatial differences and polarization trends in GIIM. Secondly, a decrease in network efficiency implies an increase in critical nodes throughout the network, leading to higher levels of GIIM spillover among provinces. This enhances the overall stability of the network, thereby reducing the relative differences in GIIM. 4.8.2. Regression results for ego-network structure Based on the characteristics of the data, estimation was performed using panel models, with reference to the results of the Hausman test to select an appropriate Table 7. OLS regression results of the whole network structure effect. Model (1) (2) Network Density 2.420**(2.66) Network Efficiency −2.357**(−2.43) Interception −2.939***(−5.58) −0.569(−1.40) R2 0.4697 9.4256 Adj.R2 0.435 0.3537 Notes: (1)*, **, *** represent at 1%, 5%, and 10% significance levels, respectively; (2) The figures in () indicate the t-values. 22 Y. LIU ET AL. random effects model. The regression results are shown in Table 8. The regression coefficients for degree centrality, betweenness centrality, and eigenvector centrality are 0.109, 0.266, and 0.130, respectively, and are positive at the 5% and 10% significance levels. This indicates that the enhancement of centrality in each province has a significant positive impact on the GIIM level. The possible reasons are as follows: First, the higher the degree centrality, the more direct relationships a province has with other provinces, resulting in a higher degree of local association and enabling provinces to acquire relevant resources more efficiently, thereby improving the GIIM level. Secondly, provinces with higher betweenness centrality can effectively control the associative effects with other provinces, thus guiding the rational allocation of resource elements and enhancing the GIIM level in the network structure region. Finally, the higher the eigenvector centrality, the closer and more influential the core inter-provincial relationships are, promoting inter-provincial communication and cooperation, and steadily improving the GIIM level. 5. QAP analysis of drivers 5.1. Variable design The formation and evolution of the GIIM spatial correlation network are attributed to the differential resource endowments of network nodes and their collective actions. Research, notably by K. Gao and Yuan (2022), has revealed that the spatial heterogeneity and regional variations in China’s green innovation are linked to geographic spatial factors. The proximity in geographical distance facilitates easier factor mobility and mutual sharing of outcomes. Extensive research has demonstrated the impact of governmental environmental regulation on green innovation in manufacturing. Manufacturing enterprises are required to allocate substantial resources to comply with mandatory environmental policies, inevitably escalating production costs and imposing a significant burden on green innovation (Y. Zhang et al., 2018), thereby influencing the direction of green innovation initiatives. However, with the increasing intensity of mandatory environmental policies, manufacturing firms are compelled to adopt more environmentally sustainable technologies. This shift enhances production efficiency and resource utilization through GIIM, consequently reducing production costs (J. Gao et al., 2023). This complex interplay accentuates the profound influence of the level of digital transformation and environmental regulatory intensity on the dynamics between sectors. Table 8. Regression results of the ego-network structure effect. Model (1) (2) (3) Degree centrality 0.109**（2.30） Betweenness centrality 0.266**（2.19） Eigenvector centrality 0.130*（1.75） Interception 0.502***（2.59）−0.196（-0.38）0.531**（2.24） Wald 5.28** 4.82** 3.06* R2 0.0509 0.0448 0.0427 Hausman statistic 0.36 0.09 1.03 FE/RE RE RE RE Notes: (1）*, **, *** represent at 1%, 5%, and 10% significance levels, respectively;（2）The figures in () indicate the t-values. JOURNAL OF APPLIED ECONOMICS 23 Gao, J., Feng, Q., Guan, T., & Zhang, W. (2023). Unlocking paths for transforming green technological innovation in manufacturing industries. Journal of Innovation & Knowledge, 8 (3), 100394. https://doi.org/10.1016/j.jik.2023.100394 Gao, K., & Yuan, Y. (2022). Spatiotemporal pattern assessment of China’s industrial green productivity and its spatial drivers: Evidence from city-level data over 2000–2017. Applied Energy, 307, 118248. https://doi.org/10.1016/j.apenergy.2021.118248 Guan, J., Zhang, J., & Yan, Y. (2015). The impact of multilevel networks on innovation. Research Policy, 44(3), 545–559. https://doi.org/10.1016/j.respol.2014.12.007 Herring, H., & Roy, R. (2007). Technological innovation, energy efficient design and the rebound effect. Technovation, 27(4), 194–203. https://doi.org/10.1016/j.technovation.2006.11.004 He, G., & Wang, Z. (2015). Energy biased technology progress and green growth transformation ——an empirical analysis based on 33 industries of China. China’s Industrial Economy, 02, 50–62. https://doi.org/10.19581/j.cnki.ciejournal.2015.02.006 Huggins, R., Johnston, A., Munday, M., & Xu, C. (2023). Competition, open innovation, and growth challenges in the semiconductor industry: The case of Europe’s clusters. Science and Public Policy, 50(3), 531–547. https://doi.org/10.1093/scipol/scad005 Kaneko, S., & Managi, S. (2004). Environmental productivity in China. Economics Bulletin, 17(2), 1–10. Karanfil, F., & Yeddir-Tamsamani, Y. (2010). Is technological change biased toward energy? A multi-sectoral analysis for the French economy. Energy Policy, 38(4), 1842–1850. https://doi. org/10.1016/j.enpol.2009.11.061 Klump, R., McAdam, P., & Willman, A. (2012). The normalized ces production function: Theory and empirics. Journal of Economic Surveys, 26(5), 769–799. https://doi.org/10.1111/j.1467-6419. 2012.00730.x Krackardt, D. (1987). QAP partialling as a test of spuriousness. Social Networks, 9(2), 171–186. https://doi.org/10.1016/0378-8733(87)90012-8 Liang, Z., Chen, J., Jiang, D., & Sun, Y. (2022). Assessment of the spatial association network of green innovation: Role of energy resources in green recovery. Resources Policy, 79, 103072. https://doi.org/10.1016/j.resourpol.2022.103072 Liu, S., Hou, P., Gao, Y., & Tan, Y. (2022). Innovation and green total factor productivity in China: A linear and nonlinear investigation. Environmental Science and Pollution Research, 29(9), 12810–12831. https://doi.org/10.1007/s11356-020-11436-1 Malecki, P. O., & Edward, J. (2008). The evolution of technologies in time and space: From national and regional to spatial innovation systems. In M. Ron (Ed.), Economy (1st ed., pp. 401– 430). Routledge. Min, S., Kim, J., & Sawng, Y.-W. (2020). The effect of innovation network size and public R&D investment on regional innovation efficiency. Technological Forecasting and Social Change, 155, 119998. https://doi.org/10.1016/j.techfore.2020.119998 Peng, J., Xiao, J., Wen, L., & Zhang, L. (2019). Energy industry investment influences total factor productivity of energy exploitation: A biased technical change analysis. Journal of Cleaner Production, 237, 117847. https://doi.org/10.1016/j.jclepro.2019.117847 Qin, L., Kirikkaleli, D., Hou, Y., Miao, X., & Tufail, M. (2021). Carbon neutrality target for G7 economies: Examining the role of environmental policy, green innovation and composite risk index. Journal of Environmental Management, 295, 113119. https://doi.org/10.1016/j.jenvman. 2021.113119 Radziwon, A., Rong, K., Lin, Y., Yu, J., & Zhang, Y. (2022). Exploring regional innovation ecosystems: An empirical study in China. In B. Catherine, B.-H. Thierry, & C. Patrick (Eds.), Innovation policies and practices within innovation ecosystems (pp. 10–34). Routledge. https:// doi.org/10.4324/9781003279464 . Rogers, M. (2004). Networks, firm size and innovation. Small Business Economics, 22(2), 141–153. https://doi.org/10.1023/B:SBEJ.0000014451.99047.69 Sims, C. A., Lovell, M. C., & Solow, R. M. (1974). Output and labor input in manufacturing. Brookings Papers on Economic Activity, 1974(3), 695–735. https://doi.org/10.2307/2534251 30 Y. LIU ET AL. SUN, Z., FAN, J., SUN, Y., & H, L. I. U. (2022). Structural characteristics and influencing factors of spatial correlation network of green science and technology innovation efficiency in China. Economic Geography, 03, 33–43. https://doi.org/10.15957/j.cnki.jjdl.2022.03.004 Wang, J., Wu, H., & Chen, Y. (2020). Made in China 2025 and manufacturing strategy decisions with reverse QFD. International Journal of Production Economics, 224, 107539. https://doi.org/ 10.1016/j.ijpe.2019.107539 Weber, W. L., & Domazlicky, B. R. (1999). Total factor productivity growth in manufacturing: A regional approach using linear programming. Regional Science and Urban Economics, 29(1), 105–122. https://doi.org/10.1016/S0166-0462(98)00013-1 Wübbeke, J., Meissner, M., Zenglein, M. J., Ives, J., & Conrad, B. (2016). Made in china 2025. Mercator Institute for China Studies Papers on China, 2(74), 4. Wu, F., Hu, H., Lin, H., & Ren, X. (2021). Enterprise digital transformation and capital market performance: Empirical evidence from stock liquidity (in Chinese). Journal of Management World, 2021(7), 130–144+10. Xuhui, L. I., & Yitao, T. A. O. (2023). Measurement, regional differences, and causes of China’ s green and low?carbon innovation development under the “dual carbon” goals. China Population, Resources & Environment, 33(1), 124–136. Yang, X., Li, X., & Zhong, C. (2019). Study on the evolution trend and influencing factors of China’s industrial directed technical change. The Journal of Quantitative & Technical Economics, 04, 101–119. https://doi.org/10.13653/j.cnki.jqte.2019.04.006 Yuan, B., & Xiang, Q. (2018). Environmental regulation, industrial innovation and green development of Chinese manufacturing: Based on an extended CDM model. JOURNAL of CLEANER PRODUCTION, 176, 895–908. https://doi.org/10.1016/j.jclepro.2017.12.034 Zenglein, M. J., & Holzmann, A. (2019). In W. Claudia (Ed.), Evolving made in China 2025 (Vol. 8, pp. 78). Germany: MERICS Papers on China. Zhang, Y., Wang, J., Xue, Y., & Yang, J. (2018). Impact of environmental regulations on green technological innovative behavior: An empirical study in China. Journal of Cleaner Production, 188, 763–773. https://doi.org/10.1016/j.jclepro.2018.04.013 Zhang, B., Yu, L., & Sun, C. (2022). How does urban environmental legislation guide the green transition of enterprises? Based on the perspective of enterprises’ green total factor productivity. Energy Economics, 110, 106032. https://doi.org/10.1016/j.eneco.2022.106032 Zhao, N., Liu, X., Pan, C., & Wang, C. (2021). The performance of green innovation: From an efficiency perspective. Socio-Economic Planning Sciences, 78, 101062. https://doi.org/10.1016/j. seps.2021.101062 Zhao, P., Lu, Z., Kou, J., & Du, J. (2023). Regional differences and convergence of green innovation efficiency in China. Journal of Environmental Management, 325, 116618. https://doi.org/10. 1016/j.jenvman.2022.116618 Zheng, W.-L., Wang, J.-W., Jiang, A.-D., Rehman Khan, S. A., Yang, X.-Q., Zhang, X., & Zhang, Z.-Y. (2021). Study on environmental performance evaluation of different linkage development types of the logistics and manufacturing industries considering the unexpected output. Journal of the Air & Waste Management Association, 71(8), 1025–1038. https://doi.org/ 10.1080/10962247.2021.1920516 Zhou, X., Yu, Y., Yang, F., & Shi, Q. (2021). Spatial-temporal heterogeneity of green innovation in China. Journal of Cleaner Production, 282, 124464. https://doi.org/10.1016/j.jclepro.2020.124464 Кузнецов, С., Турдаков, Д., Назар, Б., & Валерий, А. (2014). Social network analysis: Methods and applications. https://xueshu.baidu.com/usercenter/paper/show?paperid= 32e66bde64fbf74278db7569b85702dc&site=xueshu_se JOURNAL OF APPLIED ECONOMICS 31