Unveiling How the Digital Economy Empowers Green Productivity: Machine Learning and FsQCA Methods

Abstract

The digital economy plays a pivotal role in advancing green productivity; however, the specific configurations driving this relationship remain underexplored. Employing the TOE theoretical framework alongside k-means clustering and fuzzy-set qualitative comparative analysis (fsQCA), we systematically examine the heterogeneous pathways through which digital economy configurations enhance green productivity in China’s Beijing–Tianjin–Hebei region. The results reveal that (1) green productivity exhibits distinct temporal evolution phases and spatial distribution patterns; (2) five characteristic digital economy city clusters emerge from the clustering analysis; (3) improvements in green productivity require specific synergistic combinations of technological, organizational, and environmental factors; and (4) antecedent conditions demonstrate complex substitution patterns across different development stages. These findings offer a configurational perspective on how digital economy architectures differentially influence regional green productivity development.

1. Introduction

The current world economy needs to undergo a green transformation [1]. In light of the increasing environmental challenges, there is an urgent need for countries to adopt a green development model that aligns economic growth with ecological preservation, thereby achieving sustainable development goals. In recent years, there has been a significant rise in academic attention towards the concept of green productivity within the global academic community [2]. Enhancing green productivity is recognized as a crucial pathway for promoting sustainable development [3]. It is essential to emphasize that during the green development process, the information technology-driven digital economy has the potential to significantly mitigate environmental pressures while facilitating the synergistic advancement of both economic and ecological systems [4]. The swift advancement of digital technology and its widespread adoption not only reduce the costs associated with environmental governance but also create new intrinsic drivers for economic growth. The digital economy holds significant importance in empowering traditional industries, improving the effectiveness of resource distribution, and accelerating the green transformation process, due to its inherent characteristics of permeability, platformality, and sharing capabilities [5].

China faces a significant challenge in reconciling economic growth with environmental sustainability principles throughout its development trajectory [6]. To address this challenge, the Chinese government has implemented a series of comprehensive measures, including strengthening environmental regulations [7], fostering technological innovation [8], and promoting financial mechanisms [9]. However, persistent issues such as limited natural resources, variable consumer pricing, and elevated population density continue to intensify the deterioration of environmental conditions. Within the framework of international competition, China has become the top polluted country in the world [10]. In reaction to this circumstance, the Chinese government has articulated the theoretical proposition that “green development is the foundation of high-quality development, and new quality productive forces are inherently green productive forces.” [11]. By focusing on green productivity as the central research subject, the Chinese government can thoroughly investigate the complex relationship between productivity development and sustainable practices. Concurrently, as social integration deepens, the regional economic growth model has undergone significant transformations, framing the digital economy as a significant avenue for improving economic competitiveness [12]. In recent years, the government of China has incorporated policies related to the digital economy and green development into its key national strategies. However, existing research has yet to establish a cohesive framework that clarifies the relationship between the digital economy and high-quality green growth [5]. The questions of whether, how, and in which dimensions the digital economy influences the green productivity of Chinese cities remain to be explored further. Addressing these inquiries serves as the primary motivation for this paper.

An examination of the existing literature indicates a divergence of opinions among scholars concerning the impact of the digital economy on urban green growth. Certain researchers contend that it exerts a beneficial impact on green productivity [13,14]. When utilizing the Difference-in-Differences model, Lyu demonstrated that the digital economy contributes to the enhancement of green total factor productivity through various mechanisms. Conversely, some scholars argue that the digital economy possesses the propensity to impact sustainable development negatively [15,16]. Cheng and Zhonghua employed spatial econometric modeling and concluded that advancements in information technology have significantly exacerbated environmental pollution. Furthermore, some researchers assert that the digital economy has both positive and negative implications for green development [17,18]. For instance, Raheem and Ibrahim found that ICT has a persistent beneficial effect on emissions while exhibiting variable effects on economic growth.

The discrepancies observed in research findings may be ascribed to a variety of factors. Firstly, the majority of empirical investigations into the digital economy are conducted at the national or provincial level [19,20,21], thereby overlooking the variations in digital economy development that exist among cities, both within and across provinces. Secondly, the digital economy represents a broad and multifaceted concept that encompasses a range of multidimensional elements; thus, reliance on a unidimensional index inadequately reflects the complex content of this field, resulting in skewed empirical findings. In order to tackle these challenges, our study focuses on the Beijing–Tianjin–Hebei region, acknowledged as one of the three principal engines driving high-quality development in China. We develop a thorough evaluation system for green productivity and, from a multidimensional perspective, investigate the specific mechanisms by which the digital economy impacts the progress of green productivity. The objective is to provide relevant insights for enhancing green productivity in cities at various stages of digital economy evolution.

This research conducted systematic empirical analyses in three key steps. First, we utilize city-level panel data from the Beijing–Tianjin–Hebei region of China, covering the period from 2011 to 2022, to assess the regional level of green productivity using the TOPSIS and the entropy weight method. We further illustrate the spatiotemporal evolution characteristics using OriginPro 2025 and ArcGIS 10.8.1 software. Second, we employ an unsupervised machine learning methodology, specifically utilizing the k-means clustering algorithm to categorize cities. Third, we examine the configurational impacts of various digital economy factors on green productivity using the fsQCA methodology, extending our analysis to multi-temporal analysis to identify potential conditional substitutions. In summary, our research holds substantial academic relevance for policy formulation and regional planning. We aim to steer urban development towards a more environmentally sustainable future.

The following sections of this study are structured in the manner outlined below. Section 2 presents the research background and establishes the theoretical framework. Section 3 delineates the research design, providing a comprehensive overview of the methodology utilized, the sample characteristics, the sources of data, and the criteria for variable selection. In Section 4, an analysis of the spatial and temporal evolution of green productivity is conducted. Section 5 categorizes cities and identifies various configurations associated with the development of green productivity. In conclusion, Section 6 presents a discussion of the research findings and outlines the corresponding managerial implications. Section 7 delivers a comprehensive summary of the study’s results, discusses the limitations inherent in the article, and proposes directions for future research.

2. Literature Review and Proposition Development

2.1. Green Productivity

The fundamental principle of green productivity is rooted in the spiral rise of productivity development. The ecological environment, as an emerging factor of production, is gradually being integrated into the economic development system, similar to conventional factors such as land, labor, capital, and technology [22]. Recently, a fundamental transformation in human history has arrived, indicating that industrial civilization must transition to ecological civilization, which also underscores the originality and inevitability of the emergence of green productivity. Green productivity can be traced back to classical Western economics, which examines the correlation between resource scarcity and economic development. [23]. In 1994, the Asian Productivity Organization (APO) launched the “Special Plan for the Environment” and put forward the concept of green productivity. This concept posits that green productivity enhances both productivity and environmental performance, concurrently increasing the rate of return [24].

In China, Xiong [25] introduced the concept of green productivity from an economic perspective, asserting that “only by selecting the green path for the development of productive forces can moderate economic growth be attained while preserving ecological balance and alleviating environmental pollution.” Following this, scholars such as Meng [26] and Xue [27] further elaborated on the theoretical framework of green productivity, defining it as “the ideal integration of static benefits and dynamic development of productivity, delineating a higher level of production status and output results” [28]. Within the digital era, green productivity pertains to the use of digital technology that enhances firms’ green total factor productivity and optimizes management and production methods [29]. The continuous advancement of digital technology is essential for promoting sustainable production, enabling comprehensive carbon reduction at the source, throughout the production process, and during end-stage sequestration [30]. Digitalization and greening are the two primary themes of the current technological revolution and industrial transformation, embodying the key characteristics of new quality productive forces. To effectively raise awareness of sustainable growth and achieve optimal green production relations, integrating digital technologies with productivity is crucial [31].

2.2. Digital Economy

Extensive research suggests that the digital economy is poised to exert a range of beneficial effects on economic development [20]. Both domestically and internationally, scholars characterize the digital economy primarily by investigating its historical evolution and foundational aspects, alongside the conceptual definitions proposed by authoritative institutions. The “G20 Digital Economy Development and Cooperation Initiative,” which was established during the 2016 G20 Summit, characterizes the digital economy as a collection of economic activities that depend on digital knowledge and information as essential components of production. Chinese scholars, such as Tong and Zhang [32], in their analysis of the White Paper on China’s Digital Economy Development (2021), identified four elements of digital economy advancement: digital industrialization, industrial digitization, digital management, and data valorization. Chen et al. [33] defined the digital economy as an economic activity that primarily relies on digitized information, utilizes internet platforms as the main conduits of information, is propelled by advancements in digital technology, and is expressed through innovative models and business structures.

Academics have conducted in-depth investigations into topics such as the impact of the digital economy on common wealth [34] and low-carbon development [35,36]. Empirical research by Li and Zhao [31] indicates that the digital economy possesses the potential to enhance the efficiency of data production and foster environmentally sustainable productivity. Furthermore, Ren and Wang [37] assert that the digital economy may facilitate the allocation of production factors toward emerging industries that prioritize resource conservation and ecological protection, thereby improving the competitiveness and ecological sustainability of green industries. These findings provide substantial theoretical support and empirical evidence for further exploration of how the digital economy can facilitate advancements in green productivity.

2.3. TOE Framework

We employ the Technology–Organization–Environment (TOE) framework to synthesize pertinent factors. The TOE framework provides an analytical framework rooted in the context of technological application [38]. The main objective of this framework is to thoroughly analyze and comprehend the process of technological innovation, with a particular emphasis on the interactions among technology, organization, and environment. It has found extensive applications in domains such as the digital economy and intelligent manufacturing, with its nonlinear interactions analyzed using fsQCA [39,40,41].

Regional digital technology conditions primarily encompass infrastructural construction, education, human resources, and technological application platforms. A robust digital infrastructure can effectively integrate intra-regional resources [42] and promote the optimal allocation of these resources. Digital technology talent serves as a crucial driving force in developing new quality productive forces [43], which can dismantle technical barriers and facilitate the continuous upgrading of green transformation.

Drawing upon the extant literature, we advance the initial proposition:

Proposition 1.

Multiple configurations of technical conditions can be sufficient to generate high green productivity.

Regional digital organizational conditions mainly include organizational innovation platforms, organizational structures, and management systems, as well as funding and resource allocation. Investment in research and development (R&D) is regarded as a fundamental element in advancing green technology. Increasing R&D expenditure enhances the output of green technological innovations [44]. The digital economy transforms scientific research results through innovation platforms, facilitating a transition in economic development towards models that prioritize knowledge intensity [45].

Consequently, drawing upon the aforementioned viewpoints, we put forward the second proposition:

Proposition 2.

Multiple configurations of organizational conditions can be sufficient to generate high green productivity.

Regional digital environmental conditions primarily consist of the digital policy environment and the digital financial environment. Governments incentivize corporate green technological innovation through policy guidance, generating carbon-intensity governance effects [46]. Digital finance can enhance regional green output by mitigating the depletion and use of conventional physical resources, such as by improving payment convenience [47]. Drawing on the aforementioned findings, we have developed a theoretical modeling framework as outlined below. Hence, we put forward Proposition 3 and Proposition 4:

Proposition 3.

Multiple configurations of environmental conditions can be sufficient to generate high green productivity.

Proposition 4.

Multiple configurations of technological, organizational, and environmental conditions can be sufficient to generate high green productivity.

3. Research Design

3.1. Research Methods

3.1.1. K-Means Clustering Algorithm

This study employs the k-means clustering algorithm to categorize cities based on similar characteristics. The k-means algorithm utilizes the Euclidean distance between data points as an iterative criterion, continuously updating cluster centers to ensure convergence to a local optimum within a finite number of iterations [48]. This method effectively identifies heterogeneity in digital economy development among cities [49]. The clustering results are visualized using radar charts and scatter plots to facilitate further analysis and interpretation.

Specifically, variables were normalized to the [0,1] range using Min–Max normalization to eliminate scale differences, ensuring an equal weighting of features in distance-based clustering [50]. We then determine the optimal number of clusters using the elbow method (based on mean deviation and the sum of squared errors) and the silhouette coefficient. The ideal number of clusters is identified when the decline in the elbow plot slows and the silhouette coefficient reaches its peak, ensuring robust clustering results.

3.1.2. Entropy Weight and TOPSIS Methods

When assessing green productivity, each indicator varies in its unit of measurement, scale, and characteristics. We employ the entropy weight method in conjunction with the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) for evaluation.

The entropy weight method, grounded in information theory, objectively quantifies the informational utility of variables to determine their weights in decision-making models, minimizing subjective bias in multi-criteria evaluations. TOPSIS provides clear proximity scores relative to ideal and anti-ideal solutions, facilitating straightforward interpretation of trade-offs. Unlike DEA, which relies on input–output structures, or fuzzy AHP, which may be affected by subjective weight biases [51,52,53], TOPSIS effectively integrates both quantitative and qualitative data through normalization, making it well-suited for our mixed-type indicator system [54].

3.1.3. QCA Method

We employ the qualitative comparative analysis (QCA) approach to develop innovative explanatory solutions. All QCA investigations are configuration-oriented, as they conceptualize cases as configurations of attributes [55]. QCA facilitates the integration of extensive qualitative case information with quantitative data, enabling the identification of various configurations of variables and their combined influence on the likelihood of specific outcomes [56,57,58].

Fuzzy qualitative comparative analysis (fsQCA) allows for different levels of membership through affiliation assignments, thereby incorporating nuanced information to differentiate between case types and degrees of variation. As a result, fsQCA possesses greater content validity compared to crisp-set QCA and multi-valued QCA [59]. Therefore, we employed the fsQCA methodology. Simultaneously, we employed multi-temporal QCA in our extensibility analyses. This methodology segments the entire research duration into several intervals and conducts an individual QCA for each interval. Multi-temporal QCA enables researchers to analyze the evolution of subgroups or conditions over time by comparing them across different periods, thereby elucidating the dynamic changes in subjects (cases) associated with these subgroups [60,61].

In the empirical analysis phase, given the combined application of machine learning and QCA methods [62,63], we follow the approach proposed by Capatina [64]. First, k-means clustering is applied to classify digital economy cities into distinct types. Subsequently, the fsQCA method is utilized to examine the configurations of causal pathways within these identified city categories.

3.2. Sample and Data Sources

The data processing for this study follows a systematic procedure. Reliable data sources, including statistical yearbooks and bulletins publicly disseminated by the National Bureau of Statistics of China and other local statistical agencies, are primarily utilized. For sample selection, we consider the schedule for the issuance and implementation of central policy documents relevant to green productivity development, as well as the availability of data. Consequently, 13 prefecture-level and higher cities in the Beijing–Tianjin–Hebei region are designated as the subjects of this study. This study employs a differentiated approach to address missing values in panel data that includes various indicators. We utilize a combination of linear interpolation, smoothing techniques, and time series models to impute the missing data effectively. The above outlines our preparation for data selection and processing.

3.3. Theoretical Framework

Synthesizing the previously discussed theoretical frameworks reveals that the three components—technology, organization, and environment—exhibit significant synergistic evolutionary dynamics. Their dynamic interactions create a complex ecosystem that facilitates high green productivity. Accordingly, we propose a theoretical model of the digital economy designed to improve green productivity (see Figure 1).

3.4. Variable Measurement

3.4.1. Outcome Variable

The foundation of green productivity encompasses resource utilization efficiency, environmental protection, and various other dimensions. By effectively integrating these aspects, the indicator system proposed by Lu et al. [65] offers a more comprehensive and detailed measurement of green productivity [42]. As fundamental components of green production, the effectiveness of resource utilization holds significant importance in the process of geographical greening. Notably, the intensity of fossil energy consumption is a critical metric for delineating the structural development of regional green production. Furthermore, enhancing the degree of environmental protection, which is a primary objective of green production, can be quantitatively assessed through key indicators such as industrial emissions intensity and solid waste generation.

This particular measurement system is detailed in

Table 1. Green productivity measurement system.

Goal Level	Criterion Level	Indicator Level	Measurement Indicator	Direction
Green productivity	Resource-saving productivity	Energy intensity	Energy consumption/GDP	Negative
		Energy structure	Fossil energy consumption/GDP	Negative
	Environmentally friendly productivity	Waste emissions	Industrial solid waste emissions/GDP	Negative
		Exhaust gas emission	Industrial sulfur dioxide emissions/GDP	Negative

Energy consumption is expressed in standard coal; fossil energy consumption is characterized by natural gas supply, expressed in millions of cubic meters.

and assessed utilizing the entropy weight and TOPSIS method.

As energy consumption at the prefecture level is not publicly disclosed, this research builds upon the conclusions of Wu, et al. [66] regarding improved DMSP-OLS-like data and incorporates the methodologies for assessing energy consumption status developed by Wu, et al. [67]. By employing the linear relationship between energy consumption and nighttime illumination across 30 provincial regions, we developed a linear simulation model through regression analysis to invert and compute energy consumption at the prefecture level.

$$E_{it}=k_t{DO}_{it}$$

(1)

where $E_{it}$ is the statistical value of energy consumption in province $i$ in year $t$; $k_t$ is the coefficient for year $t$; and ${DO}_{it}$ is the sum of the gray values of all rasters in province $i$ in year $t$.

3.4.2. Antecedent Variables

The research systematically integrates the six essential elements of the digital economy into the three-dimensional TOE framework, which encompasses three dimensions: technological (including digital infrastructure and digital technical talent), organizational (comprising organizational innovation and research and development investment), and environmental (featuring government digital policy and digital finance). The six antecedent variables are evaluated and selected as shown in

Table 2. The measurement dimension and references of antecedent variables.

	Indicator Name	Measurement Dimension	References
Technology (T)	Infrastructure	Telecommunications revenue/total population	Osmundsen and Bygstad [68]; Yang and Wang [69]
		Mobile phone subscribers/total population
		Internet broadband access users/total population
	Technical talent	Computer service and software employees/end number of employees in urban units	Wang and Liu [70]
Organization (O)	Innovations	Aggregate quantity of inventions and utility model patents related to the digital economy submitted in the current year/total population	Cheng, et al. [71]; Sun, et al. [72]
	R&D investment	Expenditure on research and experimental development	Bai, et al. [73]
Environment (E)	Government policy	Aggregate keywords pertinent to the digital economy as presented in the government’s annual work report	Lei and Wang [74]; Liu, et al. [75]
	Digital finance	Data from ‘The Peking University Digital Financial Inclusion Index of China’	Guo, et al. [76]

.

4. Analysis of Spatial and Temporal Evolution

4.1. Temporal Evolution

This study utilized OriginPro 2025 data analysis software to visually represent the results of green productivity measurements in the study area from 2011 to 2022 through three-dimensional wall charts. The green productivity values for the entire sample demonstrated an upward trend over time. The increase in efficiency was significant from 2011 to 2016; however, from 2016 to 2022, improvements in green productivity slowed and exhibited a fluctuating trend. The average efficiency value over the study period exhibited an increase from 0.5301 in 2011 to 0.7014 in 2022, reflecting a cumulative gain of 32.32% (Figure 2).

The current growth of green productivity in the Beijing–Tianjin–Hebei region has significantly slowed, underscoring the existing developmental constraints, and emphasizes the urgent need for innovative advancements and alternative strategies for progress.

4.2. Spatial Evolution

This research employed ArcGIS version 10.8.1 to analyze and map the spatial development of urban green productivity levels within the Beijing–Tianjin–Hebei region over the study period. During the study period, a notable increase in green productivity was observed across the region, although this growth exhibited significant variation among different cities. Recent trends indicate that the level of green productivity has gradually diminished from urban centers, such as Shijiazhuang, to the adjacent areas, thereby highlighting a marked imbalance in green productivity growth (Figure 3).

In 2022, only four cities demonstrated green productivity levels exceeding the average value of 0.7014, accounting for 30.77% of the total cities. Among these, only Shijiazhuang and Tangshan achieved green productivity scores above 0.8. Consequently, it is essential to recognize that attaining the “dual-carbon” goal and promoting green productivity will not happen overnight.

5. Empirical Results and Analysis

5.1. Heterogeneous City Types and Characteristics

This paper initially identifies cities with diverse characteristics using the k-means clustering algorithm. The findings suggest that the downward trends of mean deviation and sum of squared errors (SSE) slowed down when five clusters were utilized. At this point, the silhouette coefficient reaches its maximum value, further corroborating the effectiveness of the clustering outcome. Moreover, an attempt to increase the number of clusters for further segmentation reveals that the sample sizes of different city clusters are polarized, indicating that merely increasing the cluster count does not effectively differentiate among city types. Consequently, based on the aforementioned analysis, we identified five categories of city groups exhibiting similar characteristics derived from the sample data of urban characteristic variables. The detailed clustering results are displayed in

Table 3. Basic information on heterogeneous city clusters.

Type	Sample Size	Representative Cities
Cluster 0	26	Baoding, Cangzhou, Chengde, Hengshui, Zhangjiakou
Cluster 1	10	Beijing
Cluster 2	54	Qinhuangdao, Tangshan, Xingtai
Cluster 3	9	Shijiazhuang
Cluster 4	57	Handan, Langfang, Tianjin

.

By analyzing radar charts (Figure 4) and comparing the multidimensional feature means of the five types of urban agglomerations with the overall feature means of the sample, it is clear that urban agglomerations demonstrate considerable heterogeneity in the advancement of the digital economy. Cluster 0 city clusters demonstrate an exceptional performance in environmental factors, as the government actively promotes the digital economy advancement through policy initiatives and a favorable digital financial landscape. However, there are notable deficiencies in organizational structure and technological infrastructure. This category of urban agglomeration is exemplified by Baoding and Cangzhou, among others, and is designated as the policy-driven city cluster. Cluster 1, which includes Beijing, significantly surpasses the overall average in several key aspects, including technological innovation, infrastructure, and talent concentration. This superiority highlights substantial advantages in advancing the digital economy, leading to its classification as a multidimensional leading city cluster. In contrast, Cluster 2 comprises city clusters that are underperforming across all dimensions, with the digital economy still in its nascent stages, as exemplified by traditional industrial cities like Qinhuangdao and Tangshan. These areas face numerous challenges related to digital transformation and are temporarily categorized as primary development city clusters. Cluster 3 is characterized by effective environmental policies and exceptional utilization of digital infrastructure, resulting in a more integrated digital ecosystem, thus earning the designation of infrastructure-driven city clusters. Finally, Cluster 4 exhibits a more equitable level of growth across all dimensions, with metrics that approximate the general average, indicating a balanced development trajectory. Consequently, it is classified as a balanced development city cluster.

5.2. Examination of the Relationship Between Multidimensional Factors and the Advancement of Green Productivity

This study examines the relationships between individual antecedent variables and their associations with the development of green productivity across five types of urban agglomerations, utilizing scatter matrix plots (Figure 5). The internal relationships within heterogeneous urban agglomerations generally exhibit positive correlations among the antecedent variables. However, significant disparities exist in the strength of these correlations. This is particularly evident in the asymmetric correlations within critical dimensions such as technological talent, innovation capacity, digital infrastructure, policy support, and the financial environment, which offer significant insights into the synergistic interactions among the various components of the digital economy.

Furthermore, a positive correlation has been identified between the antecedent variables and green productivity, providing initial evidence of the enhancing impact of the digital economy on green productivity. This relationship is characterized by a nonlinear and conditionally dependent nature, rather than a simple linear association. The extent to which the key drivers exert their influence varies significantly across different urban clusters. In response to this observation, the forthcoming study will employ the fsQCA methodology to conduct a comprehensive investigation into the diverse configurations by which the digital economy facilitates the improvement of green production across various urban contexts, as well as to clarify the patterns governing the synergistic interactions among multiple factors.

5.3. Findings from fsQCA

5.3.1. Data Calibration

This work employs fsQCA 3.0 analytic software, calibrates separately according to city clusters, and references Du and Jia [77] to utilize direct calibration in establishing three reference points for the case sample. The 95%, 50%, and 5% quantile values serve as calibrated benchmarks for fully affiliated, intersecting, and fully unaffiliated points, respectively. The antecedent conditions for certain circumstances are precisely 0.5 after calibrating, modified to 0.501 based on the existing research [57].

Table 4. Variable calibration 1.

Variable Name			Calibration Anchor Point
			Fully Affiliated
			0	1	2	3	4
Outcome variables		Green productivity	0.768	0.678	0.753	0.800	0.799
Antecedent variable	(T)	Infrastructure	0.246	0.691	0.206	0.195	0.239
		Technical talent	0.090	0.097	0.094	0.164	0.106
	(O)	Innovations	0.130	0.558	0.058	0.049	0.053
		R&D investment	549.625	2747.000	440.910	235.080	211.671
	(E)	Government policy	38.000	34.500	13.000	32.800	17.000
		Digital finance	302.071	343.558	157.323	302.442	280.338

and

Table 5. Variable calibration 2.

Variable Name	Calibration Anchor Point
	Intersection Points					Fully Unaffiliated
	0	1	2	3	4	0	1	2	3	4
GP	0.696	0.633	0.581	0.780	0.690	0.690	0.592	0.408	0.728	0.568
TI	0.145	0.515	0.065	0.180	0.137	0.137	0.461	0.030	0.123	0.082
TT	0.074	0.087	0.074	0.140	0.084	0.084	0.077	0.049	0.132	0.058
OI	0.015	0.329	0.004	0.028	0.014	0.014	0.211	0.001	0.011	0.006
OR	32.555	1725.250	19.601	130.618	37.966	37.966	1222.710	3.071	79.513	15.157
EG	22.000	14.000	4.000	28.000	11.000	11.000	5.350	0.000	13.400	3.000
ED	237.486	277.659	115.410	246.433	233.704	233.704	184.240	45.184	162.722	179.699

present the calibration findings.

5.3.2. Necessity Analysis of Individual Conditions

Prior to undertaking the sufficiency analysis, we first perform a necessity analysis, in which consistency values exceeding 0.9 indicate that the condition variable is necessary for the result variable [78]. The findings indicate that none of the antecedent variables met the consistency threshold of 0.9, except for TT across the third city cluster. This suggests that individual antecedent factors do not solely determine the high level of green productivity; rather, such effects arise from the intricate interrelations and alignments among the various conditions present in the digital economy (

Table 6. Necessary condition analysis (NCA).

Antecedent Condition	0		1		2		3		4
	Consistency	Coverage	Consistency	Coverage	Consistency	Coverage	Consistency	Coverage	Consistency	Coverage
TI	0.675	0.654	0.522	0.541	0.721	0.764	0.708	0.619	0.705	0.679
~TI	0.636	0.625	0.753	0.700	0.587	0.560	0.567	0.513	0.589	0.627
TT	0.622	0.610	0.808	0.797	0.677	0.728	0.943	0.791	0.684	0.677
~TT	0.655	0.636	0.396	0.386	0.696	0.655	0.440	0.416	0.642	0.665
OI	0.537	0.612	0.818	0.810	0.672	0.835	0.790	0.702	0.597	0.647
~OI	0.747	0.638	0.408	0.396	0.673	0.567	0.530	0.471	0.657	0.623
OR	0.508	0.605	0.812	0.816	0.689	0.850	0.783	0.749	0.643	0.753
~OR	0.769	0.635	0.392	0.375	0.654	0.553	0.570	0.473	0.689	0.613
EG	0.624	0.624	0.780	0.797	0.667	0.746	0.565	0.513	0.619	0.643
~EG	0.732	0.696	0.418	0.393	0.596	0.543	0.755	0.658	0.709	0.700
ED	0.681	0.679	0.816	0.798	0.726	0.742	0.863	0.757	0.689	0.653
~ED	0.652	0.652	0.406	0.399	0.544	0.537	0.562	0.507	0.570	0.618

).

5.3.3. Sufficiency Analysis of Configuration Fit

This paper employs a truth table algorithm to determine relevant variable combinations associated with the outcome, establishing a raw consistency threshold of 0.8, a frequency threshold of 1, and a Proportional Reduction in Inconsistency (PRI) threshold of 0.7 as the criteria for analysis [79,80]. As illustrated in Table 6, the examination of five clusters produced nine configurations that result in high levels of green productivity. The overall solution consistency exceeds 0.9, indicating that more than 90% of the case cities exhibiting these nine conditional configurations demonstrate significant advancements in green productivity. Furthermore, the overall solution coverage surpasses 0.23, suggesting that these configurations account for over 23% of instances of high green productivity development, as detailed in

Table 7. Configuration of high levels of green productivity development.

Antecedent Condition	0	1	2					3	4
	A	B	C1	C2	C3	C4	C5	D	E
Infrastructure	◯		◯		★	★	★	★	◯
Technical talent	◯	★	★	◯	⭑	◯	★	★	★
Innovations	◯	★	★	★		★	◯	★	◯
R&D investment	◯	★		★	★	★	○	★	★
Government policy	◯	★	⭑	⭑	◯		⭑	◯	★
Digital finance	◯	★	★	★	★	★	★	★	★
Consistency	0.914	0.934	0.936	0.903	0.957	0.924	0.921	0.926	0.935
Raw coverage	0.303	0.692	0.283	0.333	0.257	0.377	0.324	0.500	0.238
Unique coverage	0.303	0.692	0.006	0.004	0.012	0.036	0.058	0.500	0.238
Coverage of overall solution	0.303	0.692	0.538					0.500	0.238
Consistency of overall solution	0.914	0.934	0.924					0.926	0.935

★ or ◯ signifies a core condition; ⭑ or ○ signifies a marginal condition; ★ or ⭑ signifies the existence of a condition; ◯ or ○ signifies the lack of a condition; a blank field indicates that the condition is of little importance for the outcome.

and Figure 6.

~TI*~TT*~OI*~OR*~EG*~ED (A). In Configuration A, all identified conditions are core conditions that are absent, indicating that the digital economy is currently underutilized in enhancing green productivity within policy-driven urban agglomerations. This shortfall is primarily due to the insufficient foundational elements of the digital economy. Technically, these urban areas often lack essential capabilities in digital technological research and development, and the utilization of advanced technology, including big data and artificial intelligence, remains relatively limited, thereby hindering the comprehensive green transformation of industrial sectors. In terms of organizational innovation, firms have not adequately invested in digital transformation, and the mechanisms for collaboration among industry, academia, and research institutions are insufficient, resulting in interruptions within the continuum of green technological innovation. For instance, in Cangzhou City, despite the local government’s implementation of various initiatives, the anticipated benefits of these policies have not been fully realized due to inadequate digital infrastructure and a deficiency of skilled professionals. Currently, the green transformation efforts in these cities predominantly rely on traditional approaches, such as accommodating the relocation of industries from the Beijing–Tianjin–Hebei region and enacting administrative measures to facilitate energy restructuring. Moving forward, it is essential to prioritize integrating digital technology with sustainable production practices.

TT*OI*OR*EG*ED (B). Configuration B integrates adequate technical talent, substantial research and development (R&D) investment, robust innovation capacity, effective policy direction, and exceptional digital finance, considering these as core components. This configuration outlines a clear pathway to enhancing green productivity for cities that are leaders in the digital economy. Consider Beijing as a case study. The city has established a systematic ecosystem to advance green productivity powered by the digital economy. By leveraging the innovative strengths of the Zhong Guan Cun Science Park, Beijing has yielded significant decreases in energy usage and pollutant emissions across management, operations, and production processes. Concurrently, Beijing has witnessed consistent growth in patent applications for digital technologies, along with a substantial increase in enterprise investment in research and development, underscoring support for technological advancements in new energy, energy conservation, and environmental protection. Furthermore, Beijing has spearheaded the implementation of the Digital Economy Promotion Regulations and has integrated green development into the digital economy assessment framework. The multi-factor synergies in this configuration contribute significantly to green productivity, surpassing the national average.

The findings reveal five distinct configurations for primary developmental urban agglomerations. Urban areas with diverse resource endowments can develop unique growth models by reorganizing their factors, with a fundamental focus on achieving an optimal alignment between critical elements and local industries. Among these five configurations, digital finance emerges as a pivotal condition for existence. The underlying mechanism involves enhancing network effects within the green market via platform facilities and prioritizing financing requirements for energy conservation and environmental preservation sectors, supported by policy guidance. The initiatives undertaken by Tangshan City are particularly noteworthy. The city capitalizes on the opportunities presented by the digital transformation of the iron and steel industry, employing digital twin technology to generate the visualization of the converter steelmaking process and establishing an intelligent manufacturing system that encompasses the entire production cycle. Furthermore, Tangshan’s digital inclusive finance platform has developed a precise identification mechanism for green enterprises and projects by integrating environmental credit data, thereby improving the accuracy of green credit allocation and achieving a profound integration of digitization and decarbonization efforts.

TI*TT*OI*OR*~EG*ED (D). The D configuration signifies a Pareto improvement in green productivity, facilitated by the synergistic interaction of facility infrastructure, human resources, innovation, research and development (R&D), and digital finance, despite a relative lack of policy guidance. A robust digital infrastructure markedly improves the marginal productivity of additional inputs. The development of Shijiazhuang, which serves as a representative example of infrastructure-driven urbanization, underscores the strategic importance of digital infrastructure. The city has implemented the province’s first national-level smart city platform for spatial and temporal big data, enabling real-time monitoring and response to air, water, and solid waste pollution. Furthermore, Shijiazhuang fosters partnerships among industry, academia, and research institutions to establish a digital technology transfer center, aiming to promote the industrial application of green technologies. This configuration suggests that once digital infrastructural development meets a specific criterion, it can directly enhance the efficacy of environmental governance and increase the digital economy’s contribution to ecological growth by reducing innovation costs and optimizing resource allocation.

~TI*TT*~OI*OR*EG*ED (E). The E configuration suggests that, within the constraints of infrastructure and innovation capacity, regions can enhance green productivity through the adaptive modification of policy regulations, technological expertise, investments in R&D, and financial digitization. The underlying mechanism of this configuration is that governmental regulations establish the institutional framework necessary for such improvements. Concurrently, technical expertise enhances the effectiveness of R&D inputs, thereby partially alleviating the limitations in innovation capacity. Furthermore, financial instruments, such as digital bonds, play a crucial role in directing social capital into essential sectors. For example, Tianjin, operating under the policy framework of the Pilot Free Trade Zone, has initiated groundbreaking projects, such as the Smart Zero-Carbon Terminal, which exemplifies the successful integration of intelligent logistics and green energy. Additionally, the implementation of the Haihe Talent Plan has enabled Tianjin to attract skilled professionals in critical areas, including artificial intelligence and next-generation information technology. This configuration suggests that while deficiencies in digital infrastructure and innovation capabilities may hinder the advancement of green productivity when considered in isolation, the integration of complementary factors can nonetheless lead to significant progress in green productivity.

Overall, the results of the fsQCA indicate that each configuration sufficient to achieve high green productivity includes more than a single condition related to technology, organization, or environment. Therefore, propositions 1, 2, and 3 are not supported. Furthermore, the results show that different configurations of technology, organization, and environment can produce the same outcome, such as green productivity. This finding suggests that multiple causal pathways exist to achieve high green productivity, thereby supporting proposition 4.

5.3.4. Robustness Checks

According to the study findings of Zhang and Du [81], when minor modifications are applied to the operation, and a subset link exists among the created outcomes, the results are considered robust. Firstly, we increased the case frequency threshold from 1 to 2, followed by an elevation of the raw consistency threshold from 0.8 to 0.85. We then conducted separate fsQCA analyses. The results indicate that the configurations either remain unchanged or are included within a subset of the existing configurations, further confirming the robustness of the conclusions derived from this research to a significant extent.

5.3.5. Extensibility Analysis

Acknowledging that configurations may evolve over time, we utilize a multi-temporal QCA method, in which data from each temporal period is calibrated and examined independently. The extensibility analysis is organized into four components: data calibration, necessity analysis, sufficiency analysis, and robustness checks. The findings suggest that individual antecedent variables are not necessary for advancing green productivity, and the configurations demonstrate a degree of stability. Therefore, we will present a thorough examination of the findings derived from the sufficiency analysis.

In accordance with the structure outlined in the National Economic and Social Development Plan of the People’s Republic of China, this study delineates the temporal framework into three principal phases: the first phase corresponds to the “12th Five-Year Plan” (2011–2015), the second phase corresponds to the “13th Five-Year Plan” (2016–2020), and the initial stage of the “14th Five-Year Plan” (2021–2022) constitutes the third phase. Throughout the study periods, we identified a total of ten configurations indicative of high green productivity development. The findings are presented in

Table 8. Multi-temporal configuration table.

Antecedent Condition	2011–2015					2016– 2020	2021–2022
	A1		A2	A3	A4	B	C1	C2	C3	C4	C5
	A1a	A1b
Infrastructure		○	★	★	★	◯	○	⭑	◯	◯	★
Technical talent	◯	◯	◯	◯	★	★	○	○	○	○	★
Innovations	⭑	○		★	◯	◯	◯	◯	○	⭑	◯
R&D investment	⭑		★	★	◯	◯	○	⭑	★	★	◯
Government policy		⭑	◯	◯	★	★	○	○	⭑	○	★
Digital finance	★	★	★		◯	◯	★	★	◯	◯	◯
Consistency	0.9643	0.8748	0.9067	0.8799	0.8834	0.8086	0.8827	0.9480	0.8646	0.9159	0.9398
Raw coverage	0.2286	0.1163	0.1020	0.1303	0.0574	0.1650	0.1070	0.1227	0.0912	0.0733	0.1167
Unique coverage	0.1020	0.0385	0.0105	0.0387	0.0289	0.1650	0.0501	0.0673	0.0524	0.0142	0.0883
Coverage of overall solution	0.3565					0.1650	0.3523
Consistency of overall solution	0.9085					0.8086	0.9126

★ or ◯ signifies a core condition; ⭑ or ○ signifies a marginal condition; ★ or ⭑ signifies the existence of a condition; ◯ or ○ signifies the lack of a condition; a blank field indicates that the condition is of little importance for the outcome.

.

The evolution of multi-temporal configurations can be classified into four distinct categories: “turning trajectories,” “buffer-dominated trajectories,” “hybrid trajectories,” and “dominated trajectories” [82]. Empirical findings indicate that the individual antecedent conditions exhibit a non-stationary and alternating pattern throughout the study periods, aligning with the evolutionary nature of “hybrid trajectories.” Various antecedent factors continue to influence the evolution of regional green productivity through dynamic combinations, thereby underscoring the complexity and dynamism inherent in the advancement of regional green production.

Meanwhile, we conducted a comprehensive examination of the substitution relationships among configurations (see Figure 7). This series of images employs a three-color labeling method: green indicates antecedent conditions that converge within configurations, blue indicates the presence of conditions, while yellow signifies their absence. The analysis reveals significant differential characteristics in the alternative linkages between the configurations.

Figure 7a specifically illustrates a dual asymmetry in both the quantity and form of condition substitution. In the quantitative aspect, the alternative combinations reveal an asymmetric relationship between N and M. Regarding the nature of the conditions, those with varying attributes can yield equivalent substitution effects. Figure 7b demonstrates the substitution relationships established by a singular condition, showcasing precise substitutions among critical circumstances. Figure 7c,d exhibit similar characteristics. In specific configurations, enhanced green development performance can be achieved irrespective of the existence or nonexistence of the particular antecedent condition, exemplifying the system’s property of “different routes leading to the same destination.” Figure 7e highlights the complex interdependencies present in specific configurations, suggesting that the conditions can complement one another. The substitution relationship among the previously mentioned configurations underscores the nonlinear and context-dependent nature of the interactions among the various factors involved in the green development process.

6. Discussion and Implications

This section provides a comprehensive analysis of the aforementioned conclusions, situating them within a comparative framework by examining similar circumstances across various regions worldwide.

Section 4 of this study reveals that green productivity in the Beijing–Tianjin–Hebei region follows a growth trajectory. Specifically, green productivity experienced significant growth between 2011 and 2016, while from 2017 to 2022, the growth exhibited a fluctuating trend. This phased disparity may result from the combined effects of technology diffusion driven by initial policy initiatives—such as the Beijing–Tianjin–Hebei coordinated development strategy—and subsequent regional development imbalances. These imbalances are exemplified by the digital economy in Beijing advancing at a different pace than the transformation of traditional industries in Hebei, as well as by international technological barriers. These findings corroborate the work of Liu, et al. [83], who document the volatility and heterogeneity pervasive in global green transitions, emphasizing that uneven technological levels, inconsistent policy implementation, and complex, diverse energy structures cause imbalances between regions.

Section 5 of this study employs the fsQCA method to identify different configurations through which the digital economy promotes green productivity growth. This finding is similar to the research of Yu, et al. [84], who observed that, due to the unique economic structures and local conditions of countries involved in the One Belt One Road Initiative, the impact of the digital economy on green growth may vary. This confirms that no universal model exists for green transformation, underscoring the importance of formulating policies tailored to local contexts.

Our multi-temporal analysis revealed significant dynamic substitution relationships among digital economy conditions, indicating that the allocation of digital resources in regional development must be flexibly adapted to technological iterations and policy adjustments. This conclusion is supported by other studies as well. For example, Zhang [85] examined the impact of digital economy conditions on China’s carbon emission efficiency and found that high carbon emission efficiency could be achieved through the flexible integration of digital human resources, finance, and environmental factors. Collectively, these findings suggest that the dynamic nature of digital economy conditions requires regions not only to establish high-frequency monitoring systems but also to develop flexible policy frameworks capable of adapting to the unpredictability of technological substitution.

The policy implications derived from this study are outlined as follows:

Governments should implement phased management strategies by establishing dynamic policy frameworks that adapt to different stages of growth. For example, to effectively leverage digital technologies during periods of accelerated green productivity growth, government bodies should prioritize integrating artificial intelligence and big data analytics into clean energy infrastructure and intelligent manufacturing systems. During phases of stagnation, the strategic focus should shift toward adopting blockchain-enabled financing mechanisms and Internet of Things (IoT)-based resource tracking to overcome existing developmental constraints. Simultaneously, establishing a dynamic national monitoring framework is essential to facilitate coordinated progress across all stages of development.

Tailored strategies should align with the specific clusters within the digital economy. To capitalize on the benefits of the digital economy, policy-driven cities should enhance their strategic planning by developing specialized frameworks tailored to the digital economy and establishing governance structures that integrate digital technologies with environmentally sustainable industries. Multidimensional leading clusters must integrate advanced digital technologies into green manufacturing processes. Primary development clusters should leverage their local resource endowments to promote the advancement of digital inclusive finance, utilizing tools such as green credit, big data risk management platforms, and carbon accounting systems to channel financial resources toward clean energy initiatives. Infrastructure-driven areas are encouraged to capitalize on emerging infrastructures, including 5G base stations and big data centers, to implement intelligent energy management systems that facilitate precise scheduling and optimize the efficiency of renewable energy sources such as solar and wind power. Balanced development clusters should harness their institutional strengths to establish collaborative innovation centers that bring together industry, academia, research, and application sectors, thereby addressing the critical challenge of limited green technology innovation capacity. Crucially, government agencies must coordinate digital and green transitions to ensure synergistic benefits rather than isolated improvements.

The advancement of systemic management cognition is essential. Local governments are encouraged to evaluate the synergistic effects of the digital economy and to develop a governance framework that incorporates the “technology–organization–environment” triad when formulating regional strategies for green development. By employing a systematic planning methodology, regions can effectively address the prevalent challenge of “fragmented” governance that characterizes conventional decision-making processes, thereby achieving Pareto optimal outcomes in both economic advancement and ecological preservation.

7. Conclusions

The research conclusion of this study has three main aspects:

First, the progression of green productivity within the Beijing–Tianjin–Hebei region demonstrates distinct characteristics that vary across different temporal stages. Specifically, there was a period of rapid growth from 2011 to 2016, which was followed by a deceleration and a pattern of fluctuations from 2016 to 2022. In terms of spatial distribution, the development is characterized by an uneven, multi-centered structure.

Second, the Beijing–Tianjin–Hebei region is classified into five distinct types of city clusters based on the dimensions of the digital economy: policy-driven, multidimensional leading, primary development, infrastructure-driven, and balanced development clusters. Subsequent configuration analyses indicate that the advancement of green productivity is not solely reliant on a single aspect of the digital economy; rather, it emerges from the synergistic interaction of various factors, with significant variations observed across the different cluster configurations.

Third, a multi-temporal longitudinal analysis reveals that the individual antecedent conditions influencing the development of green productivity driven by the digital economy exhibit characteristics of hybrid trajectory evolution. Four distinct types of substitution relationships can be identified among these configurations: asymmetric substitution, characterized by the unequal nature and quantity of combinations of antecedent conditions; precise substitution, marked by the exact exchange of conditions; multiple paths, indicated by the coexistence and absence of factors that collectively generate a high level of configuration; and cross-dependence, evidenced by the synergistic interactions among the antecedent conditions.

This study has two primary limitations. First, in measuring the energy structure, the analysis relied solely on the volume of natural gas supply as a proxy variable due to the lack of comprehensive statistical data. This approach excludes other critical fossil fuel consumption metrics, such as coal and oil, which may compromise the representativeness of the energy structure assessment. Second, although the outcome variable system includes dimensions such as energy intensity, energy structure, and waste emissions, it may overlook essential factors influencing variations in regional green productivity. Future research should develop a comprehensive energy structure indicator system based on improved data availability, integrating the proportions of fossil fuels, clean energy metrics, energy conversion efficiency, and characteristics of energy transmission and distribution. Furthermore, greater emphasis should be placed on measuring outcome variables.