How Can Data Elements Empower the Improvement of Total Factor Productivity in Forestry Ecology?—Evidence from China’s National-Level Comprehensive Big Data Pilot Zones

Abstract

In the context of global climate change and the deepening of ecological civilization construction, forestry, as an ecological security barrier and green economic engine, faces many challenges to the enhancement of its ecological total factor productivity in the traditional development model. As a new type of production factor, the data factor provides a new path to crack the bottleneck of forestry eco-efficiency improvement. Based on China’s provincial annual panel data from 2014 to 2022, this study systematically examines the impact and mechanism of data factors on forestry ecological total factor productivity by using the SBM-GML model and dual machine learning model. It was found that data factors have a significant contribution to forestry ecological total factor productivity, a conclusion that passes a series of robustness tests and endogeneity tests. The analysis of the mechanism shows that the data factor enhances the total factor productivity of forestry ecology mainly through three paths: promoting the progress of forestry technology and promoting the rationalization and advanced structure of the forestry industry. Further analysis showed that the promotional effect of data elements is more obvious in regions with a high level of green finance development, high intensity of environmental regulation, and strong financial autonomy. It is recommended to systematically promote the in-depth application of data elements in forestry, build a data element-driven innovation system for the whole chain of forestry, and implement regionally differentiated data element-enabling strategies.

1. Introduction

Forestry, as both a green economic engine and an ecological security barrier, plays a vital role in achieving carbon peaking and neutrality, preserving biodiversity, and promoting sustainable development [1,2]. Ecological total factor productivity (TFP) serves as a key indicator of high-quality forestry development. It reflects not only the efficiency of capital and labor input allocation but also the capacity for ecological resource transformation, the internalization of environmental externalities, and the realization of ecological product value. Currently, forestry faces mounting challenges amid global climate change and the push for ecological civilization. Traditional factor-driven growth models are yielding diminishing marginal returns, straining ecological carrying capacities, and exacerbating value loss through product homogeneity [3]. Simultaneously, low-end industrial chain positioning, slow diffusion of smart forestry technologies, and regional development imbalances constrain productivity improvements [4].

In response, the Third Plenary Session of the 20th CPC Central Committee highlighted the imperative to “significantly increase total factor productivity” and “develop a national integrated technology and data market”. Data elements are thus positioned as strategic enablers of technological upgrading, industrial restructuring, and system-wide synergy. To overcome the ecological–economic dichotomy, China must shift from “factor accumulation” to “intelligent synergy”, strengthen its technological foundation [5], advance industrial structure [6], and improve institutional mechanisms such as ecological compensation and market-based trading [7]. These coordinated efforts aim to optimize forestry’s ecological, economic, and social functions, supporting both the “two mountains” philosophy and long-term sustainability.

The concept of ecological total factor productivity in forestry builds upon the traditional TFP theory by integrating ecological factors, environmental constraints, and the valuation of ecosystem services [8]. It captures the overall efficiency of the “ecological–economic–social” system in converting inputs into both economic and ecological outputs under resource limitations. Conceptually, green TFP emphasizes the synergistic optimization of ecological and economic benefits within the constraints of land, labor, and capital, thus going beyond the scope of single-output productivity measures. Methodologically, recent studies have shifted from radial distance functions to non-radial models such as SBM and EBM, which better accommodate forestry’s multi-input–multi-output characteristics. For example, a study introduces the slack-based SBM approach to improve efficiency measurement under undesirable outputs [9]; another proposes the EBM model that integrates radial and non-radial features for more accurate frontier estimation [10]; and another applies these methods specifically to the forestry sector, demonstrating their effectiveness in ecological performance evaluation [11]. Indicator systems typically include inputs such as land, labor, capital, and energy, along with economic and ecological outputs. Forestry-related pollution emissions—the “three wastes”—are also included as undesirable outputs [12,13]. Scholars have developed a systematic analytical framework for examining ecological TFP in forestry, focusing on three key drivers: the institutional environment, technological innovation, and industrial restructuring. At the policy level, regulatory tools such as ecological redlines and logging quotas ensure ecological thresholds are respected [14], while incentive mechanisms like carbon trading and eco-compensation help internalize environmental externalities through market signals [15]. On the technological front, precision equipment and digital tools (e.g., drones, blockchain, remote sensing) reduce ecological losses and improve input efficiency by mitigating information asymmetries [16]. Meanwhile, biotechnology and recycling technologies promote a transition from linear to circular forestry production models, although high transformation costs and adoption barriers remain among forest operators [17]. Finally, upgrading the forestry industry toward services such as eco-tourism and carbon finance helps ease ecological pressures, while non-timber products gain added value through branding and eco-certification [18].

Data elements have emerged as a new core production factor—alongside land, labor, and capital—redefining value creation logic and industrial development paradigms in the context of rapid global digital transformation and ecological civilization construction. Enabled by digital technologies and infrastructure, data elements now facilitate efficient collection, circulation, and value conversion. By transcending industrial boundaries, they optimize decision-making and stimulate innovation, offering a novel solution to development constraints imposed by resource and environmental limitations [19]. In forestry, data elements carry strategic significance. First, the inherent complexity and spatial heterogeneity of forest ecosystems necessitate data-driven precision governance to replace traditional empirical management and address global issues such as biodiversity loss and climate change [20]. Second, forestry’s ecological TFP remains hindered by sluggish technological progress, inefficient resource allocation, and weak value realization mechanisms [8]. The contribution of data elements can be summarized in four dimensions. (1) They enhance information flow, enabling scientific decision-making in forestry through tools such as remote sensing, big data analytics, and IoT applications for real-time resource monitoring [21]. (2) They drive technological innovation in areas like environmental surveillance, smart forestry, and precision agriculture, thereby improving both productivity and ecological outcomes [22]. (3) By offering comprehensive insights into supply and demand, they support the structural upgrading of the forestry sector, advancing modernization and sustainability [23]. (4) They improve ecological efficiency by supporting carbon sink management, ecosystem restoration, and data-informed ecological function enhancement [24].

In conclusion, data elements—arising from the deep integration of digital technology and forestry—are not merely byproducts but serve as vital drivers of forestry’s green transformation. By reshaping production functions, enhancing system coordination, and fostering new value paradigms, data elements provide sustained momentum for improving ecological total factor productivity. This approach not only reflects a modernized Chinese pathway for sustainable forestry development but also aligns with the national strategy of integrating digital and green growth. While existing research on forestry TFP has made significant strides—particularly in exploring the synergy among policies, technologies, and industrial restructuring—most studies remain narrowly scoped and overlook the systematic role of data elements. As the digital economy expands, data have increasingly emerged as a fundamental production factor, playing a pivotal role in boosting ecological efficiency and enabling green development in the forestry sector.

To enhance the ecological total factor productivity (TFP) of forestry, this study constructs an analytical framework of “data factors–transmission mechanisms–ecological efficiency” to explore the internal logic and realization paths of data-driven improvement. Empirical analysis is conducted using panel data from 30 Chinese provinces, autonomous regions, and municipalities from 2014 to 2022. A dual machine learning model is employed to evaluate the impact of data elements on forestry ecological TFP and to reveal underlying transmission mechanisms. Based on the results, policy recommendations are proposed to accelerate green transformation in the sector.

This study makes three key contributions. First, it identifies data elements as a new core production factor in forestry, breaking through the conventional focus on land, labor, and capital. In doing so, it addresses a gap in the literature by emphasizing how data elements enhance the ecological, economic, and social performance of forestry through technological innovation and industrial upgrading. Second, it introduces the “data-driven–system restructuring–ecological value-added” framework to clarify the multifaceted role of data in improving ecological TFP. This framework captures not only the input characteristics of data but also its enabling effects on technical advancement, ecological efficiency, and structural optimization. Third, the study expands the theoretical foundation of green TFP in forestry by incorporating concepts from ecological economics, resource constraints, and ecosystem service valuation into performance evaluation systems. Methodologically, the study innovatively applies a dual machine learning model that accommodates the high dimensionality and complexity of forestry data more flexibly than traditional regressions. By integrating statistical learning with causal inference logic, the model identifies how data factors affect ecological productivity through channels such as technology diffusion and industrial structure change. This approach strengthens both empirical accuracy and theoretical applicability, offering a robust basis for policy design in support of sustainable forestry development.

2. Theoretical Analysis and Research Hypotheses

By rearranging the production function and maximizing the synergy of the forestry ecological-economic system, the digital transformation powered by data elements has emerged as a key means of overcoming ecological constraints and raising the overall factor productivity of forestry ecology, in contrast to the unsustainable development path to digital transformation that depends on the use of natural resources. Through the triple transmission mechanism of value invention, structural optimization, and technological empowerment, data factor—the fundamental production component of the digital economy—has created the transformation paradigm of “data-driven-system reconstruction-ecological value-added”. The development of this analytical framework offers a methodical way to achieve the mutually beneficial evolution of ecological preservation and industrial upgrading, in addition to exposing the underlying logic of the data factor enabling forestry ecological transformation.

2.1. Data Elements, Technological Progress in Forestry, and Total Factor Productivity in Forestry Ecology

By rearranging the logic of technology generation, optimizing the innovation ecosystem, and enhancing the effectiveness of technology diffusion, the methodical embedding of data elements has created a multifaceted driving system for the advancement of forestry technology. This system encourages the development of the forestry technology system in the direction of intelligence. Data resources are a new kind of production factor whose main purpose is to create a closed-loop innovation chain of “data insight-technology research and development-practice verification” by overcoming the information and time constraints of traditional technology research and development. The targeting and predictability of forestry science and technology innovation are greatly enhanced by the thorough integration of data elements during the technology research and development process [25]. Forestry researchers can create digital models that span the full life cycle by using the Big Data Pilot Zone’s data aggregation platform to analyze multi-dimensional data in real time, including soil microenvironmental changes, forest growth dynamics, and climate response mechanisms. By accurately determining the best combination of cultivation parameters under various land conditions and by using simulation and deduction to predict the ecological effects of technology application, these intelligent algorithms based on massive data training are causing a paradigm shift away from empirical trial and error and toward data-driven technology application. The forestry production system has been intelligently reconfigured at the technological application level thanks to data components. A full-area sensing network for silvicultural planning, nursery management, and harvesting operations has been created as a result of the close integration of IoT nodes and remote sensing grids. Through edge computing, the dynamic data flow that results optimizes operational decisions in real time. In addition to increasing the accuracy of technology implementation, this closed-loop control system of “sensing-analysis-execution” also feeds technology iteration through the ongoing accumulation of real-world data and encourages the slow maturation of technologies like intelligent logging equipment and unmanned aerial vehicle ecological monitoring [26]. The data element’s ability to overcome the island effect of traditional technology dispersion is particularly important. Key technologies like seed selection and breeding, ecological restoration, and others have been able to rapidly spread across administrative boundaries, particularly in remote forested areas, thanks to the blockchain-based forestry technology traceability system and cross-regional knowledge-sharing cloud platform. This has created an empowering mechanism for technological leapfrogging. By drastically lowering the adoption threshold, this networked diffusion mode enables green technologies like biodiversity conservation and water-saving farming to spread widely throughout the nation and establishes the technological groundwork for the development of sustainable forestry [27].

By redefining the combination of ecological factors of production and optimizing the transformation route of the ecological economy, technical advancements in forestry have emerged as the primary driver of forestry ecology’s increased total factor productivity. Its main mechanism is the use of technical innovation to realize the value leap of ecological resources and system optimization, as well as to overcome the law of declining marginal gains of traditional factors of production. Technological advancements have recreated the intensive path of natural resource utilization in the factor allocation dimension [28]. The use of intelligent sensing systems allows for qualitative changes in the regulation of ecological factors like light and water, enables dynamic adaptation and accurate matching of resource inputs, and dramatically lowers resource loss while preserving the growth efficiency of forest trees. Through the closed-loop design of the material cycle, the external negative effect is converted into the production factors within the system, and the new slow-release technology’s application greatly increases the efficiency of nutrient supply. This technology-induced factor substitution effect essentially alters the crude production mode [29]. Technological innovation has built a full-cycle ecological risk prevention and control system at the process control level. The establishment of an integrated air and sky monitoring network has significantly reduced the depreciation of ecological service functions through preventive intervention, revolutionizing the early warning response mechanism for ecological threats like the likelihood of forest fires and the routes by which pests and diseases spread [30]. Advances in biomass conversion technologies have changed how forestry waste is treated, turning conventional environmental burdens into materials for soil improvement or bioenergy and creating a valuable avenue for material recycling. Technological advancements have broadened the multifaceted value space of ecological products in terms of value production. A multiplier effect has been created by the clever upgrading of the composite operation mode, which has combined economic benefits with ecological efficacy, such as carbon sink capacity and species conservation per unit of forest land. By precisely estimating the value of forest ecological services, the development of carbon measuring technology has further expanded the marketization channel of ecological capital and activated the potential resources [31]. The optimization of the factor productivity structure reflects deeper changes, and the widespread use of automated equipment not only increases the labor factors’ marginal output but also lessens the ecological footprint by adjusting the energy consumption structure fundamentally. In essence, this trend of technological advancement has facilitated the shift in forestry development from “scale expansion” to “quality leap”, and it has given the economy and ecology a long-lasting boost. In light of this, the following hypothesis is proposed for this study:

H1: 

The data element contributes to the ecological total factor productivity of forestry by improving technological progress in forestry.

2.2. Data Elements, Rationalization of Forestry Industry Structure, and Total Factor Productivity in Forestry Ecology

By fixing factor mismatches, reshaping the degree of industrial relevance, and optimizing the value distribution mechanism, the forestry industry system is encouraged to form a rationalized pattern of supply and demand matching, regional coordination, and hierarchical balance. The thorough integration of data elements provides the underlying reconstruction power for the rationalization of the forestry industry structure. Cracking the structural contradiction of traditional forestry production and creating a dynamic equilibrium mechanism of “data flow guiding factor flow-factor flow driving structural adjustment” are the fundamental values of data elements, which serve as a “digital lever” for industrial structural adjustment. Data components enable precise matching and dynamic adaptation of industrial factors in the factor allocation dimension [23]. Forestry entities are able to create a factor allocation decision-making model that encompasses the entire chain of forest management, processing, and services by utilizing the big data pilot zone’s industry monitoring platform to analyze the carrying capacity of regional resources, market demand elasticity, and technology diffusion gradient in real time. Under the traditional model, this data-driven factor rebalancing mechanism successfully corrects the mismatch between excess processing capacity and insufficient ecological service supply. For instance, it tracks the capacity utilization rate to guide the exit of inefficient wood-processing equipment while increasing factor inputs into green segments like the under-forest economy and carbon-sinking forest management [32]. Data components enhance the industrial chain’s complementarity and synergy at the level of industrial associations. The blockchain-based supply chain management system facilitates the conversion of raw products into high-value-added segments, breaks down information barriers upstream and downstream, and permits deep processing, forest cultivation, and seedling breeding to create a demand-oriented linkage mechanism [33]. More significantly, the data elements have led to adaptive changes in the structure of the industry. Through real-time matching of supply and demand information, the industrial Internet platform encourages small and medium-sized businesses (SMEs) to move from uniform competition to specialized division of labor and cooperation. For instance, creating distinctive industrial clusters and creating a complementary and varied industrial layout based on the area resource endowment information. Data elements support the profound evolution of the three industries’ integration in terms of hierarchical structure. A nested development model between forestry and cultural tourism, healthcare, and other service industries has been formed as a result of the new ways that multi-source data fusion analysis has shown to realize the value of ecological products. For instance, the structural balance of material production and service supply is promoted by the design of product combinations for forest experiences that are based on data mining of visitor behavior. Through data penetration, this industrial hierarchy rebalancing process effectively removes the rigidity of industrial boundaries, creates a three-dimensional industrial structure of “ecological base-material production-service value-added”, and establishes a digital basis for the industrial system’s rationalization [34].

By altering the ecological economic system’s value transformation mechanism and factor allocation efficiency, rationalizing the forestry industry structure has emerged as a crucial strategy for raising the overall factor productivity of forestry ecology [35,36]. Its fundamental mechanism is the realization of the Pareto improvement of ecological elements and the multiplier effect of value creation through structural adjustment, which breaks through the systemic imbalance of the classic growth model. The rationalized structure unleashes the synergistic value-added potential of ecological factors at the factor allocation level [37]. Increased carbon sink output per unit of resource is positively correlated with biodiversity conservation when the processing chain is intensified, reallocating land and water resources to the field of ecological conservation. Through technological integration, the forest industry is undergoing a cleaner transformation that essentially alters the linear metabolic model of “high inputs-high losses” by converting production emissions into raw material inputs to the circular economic system [38]. The rationalized structure has created a multifaceted value realization channel for ecological products in the value transformation dimension. Promoted by the thorough integration of the three industries, the “Forestry+” model allows the same input of ecological elements to simultaneously produce numerous value outputs, including material products, cultural services, and ecological management. For instance, by mining data on visitors’ willingness to pay for ecological services, the forest recreation base effectively converts the value of ecological services into revenue. The reconstruction of the value distribution mechanism reflects deeper changes: the digital platform-supported precision marketing system has shortened the ecological product premium transfer chain, allowing forest producers to directly receive ecological added-value income, creating a closed-loop incentive known as “value creation-value realization-value feedback”. The “value creation-value realization-value feedback” closed-loop incentive is established. The rationalized structure improves the ecological economic system’s stability and sustainability in terms of system effectiveness [39]. By improving resource appropriateness, the regional specialized division of labor system lessens the intensity of ecological development. For instance, arid regions prioritize the growth of water-saving forest and grass industries, and the risk-hedging mechanism created by the combination of diverse industries successfully curbs the impulse of overcutting brought on by price fluctuations. By creating an adaptive adjustment mechanism, this close coupling of industrial structure and ecosystem not only improves the ecological total factor productivity’s static level but also fortifies its capacity for dynamic evolution, offering structural support for the long-term growth of forestry. This led to the formulation of the study’s hypothesis:

H2: 

Data elements promote the total factor productivity of forestry ecology by improving the rationalization of the structure of the forestry industry.

2.3. Data Elements, Advanced Forestry Industry Structure, and Ecological Total Factor Productivity in Forestry

Through the reconstruction of the industrial technology base, the creation of knowledge-intensive business forms, and the reshaping of the global value chain, the strategic penetration of data elements has brought innovative leaping kinetic energy into the advanced forestry industry structure and encouraged the forestry industry system to evolve in the direction of higher-order technology, service-led, and high-end value. The fundamental purpose of data elements, the “digital engine” of industrial upgrading, is to construct the “data intelligence-knowledge creation-value fission” path of industrial upgrading by overcoming the endowment restrictions of conventional factors of production. As a path to industrial upgrading, data components propel the forestry production function’s technological level to soar at the level of technological basis reconstruction. The machine learning-based forest growth prediction model shortens the breeding and species selection cycle and eliminates the trial-and-error bottleneck in biotechnology research and development; the digital twin technology-built virtual forest system encourages the shift from experience-dependent to knowledge-intensive forest management by simulating the ecological and economic impacts of various management strategies in real time [40]. The proportion of advanced aspects like research and development, design, and digital services has steadily increased as a result of this qualitative shift in the technological paradigm, which has also profoundly changed the forestry industry’s technological makeup. Data components have sparked the upgrading of the global value chain for forestry in the value chain reshaping dimension. We have examined the evolution of global technical standards for forest products through cross-border data flow and advised businesses to move from exporting low-end raw materials to developing biobased new materials, drugs derived from forests, and other high-tech products in order to gain a competitive edge at both ends of the “smile curve.” More significantly, the emergence of digital service products like carbon sink measurement platforms and ecological value assessment algorithm packages has led to the fissile development of the forestry productive service industry. This has changed the forestry industry from producing materials to a “technical standard output + data service trade.” The production of materials has given way to the “technical standard output + data service trade” in the forest sector. A novel pattern of “technology-ecology-finance” integration in forestry has been made possible by data elements in terms of industrial innovation. New business models like carbon asset securitization and ecological compensation derivatives have emerged as a result of the combination of forest carbon sink remote sensing monitoring data and green financial instruments; precision forestry, synthetic biomanufacturing, and other cutting-edge industries have emerged as a result of the interaction between forest genetic databases and bio-economic innovations [41]. This advanced industrial structure’s main goal is to create a multi-dimensional, non-linear leap upgrading mode by breaking the linear path of traditional industrial upgrading through the multiplier effect of data elements.

By altering the ecological economic system’s value creation paradigm and technology diffusion mechanism, the forestry industry’s sophisticated structure has turned into a strategic fulcrum for increasing the sector’s total factor productivity. Its main method is to overcome the conventional path dependence of productivity enhancement and achieve a leap in development quality by combining ecological value innovation with the accumulation of knowledge capital. The advanced structure encourages the knowledge-based transformation of ecological production elements at the key quality state level. The creation of a digital germplasm resource bank increases the effectiveness of using forest genetic resources, the adoption of bioinformatics technology allows biomass production per unit of forest land to surpass the limit of natural growth, and the profound integration of these technological components essentially alters the ecological factors’ output elasticity [42]. The sophisticated structure has resulted in a geometric increase in ecological value in the value generation dimension. The development of a global trading platform for forest ecological services through the assetization of satellite remote sensing data, which converts regional ecological benefits into digital assets that can be transferred across borders, and the use of forest metabolomics technology, which has increased the economic output of individual forest trees and unlocked the commercial value of secondary metabolites at the molecular level, are just two examples of the qualitative changes in the mechanism of ecological product value realization brought about by the growth of the forestry digital services industry [41]. Additionally, the sophisticated structure creates a networked spillover of environmentally beneficial innovation. Through data sharing mechanisms, the industrial innovation consortium speeds up the spread of green technologies. For instance, the blockchain-enabled industry-university-research collaborative platform allows the customized application of anti-resistant tree cultivation technology in various eco-regions, and the modularized design of intelligent logging equipment allows remote forest areas to quickly benefit from the technology spillover dividend. The deeper shift is evident in the way the productivity enhancement mechanism has changed: the addition of digital service components gradually eliminates the ecological total factor productivity’s reliance on material components, and the interaction of ecological positive externalities and the knowledge spillover effect creates a self-reinforcing enhancement mechanism [29]. Through the creation of a technology-institution-market linkage innovation mechanism, this profoundly synergistic evolution of industrial structure and ecosystem not only reconfigures the logic of generating ecological total factor productivity but also creates a higher-order path for sustainable forestry development. In light of this, the following hypothesis is put up for this study:

H3: 

The data element contributes to the total factor productivity of forestry ecology by improving the advanced structure of the forestry industry.

3. Variable Selection

3.1. Independent Variable

The independent variable of this article is the pilot policy on data elements. The Chinese government’s strategic plan for the growth of the digital economy and big data industry primarily provides the policy background for national-level big data comprehensive pilot zones. Through policy guidance and the establishment of pilot zones, the government hopes to encourage the marketization of data elements, the digital transformation of industries, and the synergistic development of regions. The development of big data was raised to a national strategy for the first time in 2015 when the State Council published the Outline of Actions for Promoting the Development of Big Data. By clearly defining specific goals and policy directions for fostering the growth of the big data industry and investigating methodical reforms in data resource management, industrial development, and application innovation, the Outline’s release demonstrated China’s high regard for the significance and strategic value of big data technology. Guizhou was authorized as the first nationwide big data comprehensive pilot zone in 2016 by the Central Internet Information Office (CIIO), the Ministry of Industry and Information Technology (MIIT), and the National Development and Reform Commission (NDRC). This action represented a significant step in the execution of China’s big data plan and offered crucial policy support to encourage the growth of the big data sector. In order to further encourage the synergistic innovation of big data applications and industrial development in different regions, the pilot zone’s scope has since been progressively expanded to include Beijing-Tianjin-Hebei, the Pearl River Delta, Shanghai, Henan, Chongqing, Shenyang, Inner Mongolia, and other significant areas. In order to select the national-level big data comprehensive experimental zones that were approved by the National Development and Reform Commission between 2016 and 2022, this study combines the time of selection to create a policy dummy variable as a measure of datasin, citing Sun Bowen [43]. The primary reason for the pilot regions’ selection is their exceptional big data application development achievement, which is common and beneficial for promotion. Thus, if a region is accepted as a pilot area between 2016 and 2022, the variable is given a value of 1 for the year following the approval and the year following, and a value of 0 for the year prior to the establishment. This is how the study sets them up. All years are given the value 0 for territories that have not established a pilot area.

3.2. Dependent Variable

The total factor productivity of forestry ecology, a comprehensive index that gauges the economic output, ecological benefits, and environmental pollution performance attained based on inputs of production factors like land, labor, capital, and energy in forestry, serves as the study’s explanatory variable. The factors of land, manpower, capital, and energy were chosen as input indicators based on the research of Chen C et al. and Yan J et al. [8,44]. Economic benefits, ecological benefits, forestry SO₂ emissions, wastewater discharges, and solid waste discharges were chosen as output indicators, and FETFP was calculated using the SBM-GML model. The indicators were set up as

Table 1. Components of an ecological total factor productivity indicator for forestry.

Norm	Indicator Name	Representation
Input element	Land input	Forest land area
	Manpower inputs	Number of employees in the forestry system
	Capital investment	Investment in fixed assets in forestry completed
	Energy inputs	Energy consumption in gross forest product
Expected outputs	Economic benefit	Gross Forestry Product
	Ecological benefit	Forestry ecosystem services
Unexpected outputs	Forestry exhaust outputs	Industrial SO2 emissions * forestry secondary production/total industrial production value
	Forestry solid waste output	Industrial solid waste generation * forestry secondary production value/total industrial production value
	Forestry wastewater output	Industrial wastewater discharges * forestry secondary production/total industrial production value

. These include the division of forestry ecosystem service areas into four first-level service systems for supply, support, cultural, and regulatory services, the application of the equivalent factor method, and the adjustment of the correction coefficient of net food profit to be measured. These forestry ecosystem services are based on the research of Xie Gao Di et al. [45]. The indirect method was used to measure the “three wastes” emissions from forestry operations. Table 1 displays the definitions of the particular variables.

The SBM-GML model is set up as follows:

$$P=min{\displaystyle \frac{1-\frac{1}{G}\sum _{g=1}^G\frac{A_g^x}{x_g^k}}{1+\frac{1}{Z+J}\left(\sum _{z=1}^Z\frac{A_z^y}{y_z^k}+\sum _{i=1}^I\frac{A_i^b}{B_i^k}\right)}}$$

(1)

$$A.t.\left\{X_g^k=\sum _{k=1}^K{r}_k{x}_g^k+A_g^x,y_z^k=\sum _{k=1}^K{r}_k{y}_z^k-A_z^k,B_i^k=\sum _{k=1}^K{r}_k{x}_i^k+A_i^b\right\}$$

(2)

$$ML={\left({\displaystyle \frac{D^t\left(x^{t+1},y^{t+1},u^{t+1}\right)}{D^t\left(x^t,y^t,u^t\right)}}\times {\displaystyle \frac{D^{t+1}\left(x^{t+1},y^{t+1},u^{t+1}\right)}{D^{t+1}\left(x^t,y^t,u^t\right)}}\right)}^{\frac{1}{2}}=EC\times TC$$

(3)

where P = 1 indicates that the decision unit is in the efficiency frontier plane and that technological efficiency has advanced; when P < 1 indicates that there is an optimization potential in the dimension of factor inputs or outputs. G stands for decision unit inputs, J for desired outputs, and Z for non-desired outputs. The forestry ecological total factor productivity (GTFP) is broken down into two dimensions by the productivity index based on the SBM directional distance function: the technical efficiency index (EC) and the technical progress index (TC). The growth of total factor productivity in forestry ecology, the improvement of environmental efficiency level, and the advancement of green technology are all indicated by measured values of the GTFP index, EC index, or TC index that are greater than 1. Conversely, a value below 1 indicates technological regression or efficiency decline.

3.3. Mechanism Variables

The goal of this study is to examine the level of total factor productivity of forestry ecology using data from three different paths: advanced forestry industry structure, forestry technological advancement, and forestry industry structure rationalization.

Because patents are the product of technological innovations and reflect the level of research and development (R&D) and innovation capacity of a region or a country in the field of forestry, Mei Yuntian et al. [46] conducted a study on the measurement of forestry technology progress indicators using the logarithm of the number of forestry patents. Patents quantify the impact of R&D expenditure and technology dissemination in addition to reflecting the market potential and usability of technology. A rise in forestry patents can help advance industrialization, technological application, and overall industry development. The industry gains new competitiveness as a result of the transformation and spread of patented technologies, which also speed up the modernization and environmentally friendly development of the forestry production mode. As a result, the quantity of forestry patents is a significant and obvious measure of technological advancement.

The reference Kong Fanbin et al. [47] study used the total output value of forestry, the sum of the output values of the secondary and tertiary industries, and the output value of forestry to measure the advanced forestry industry structure index. Because it reflects the extent to the degree to which forestry has evolved from a conventional resource-exploitation sector to an ecological and high-value technical service sector. The processing and manufacturing aspects of forestry are typically included in the secondary industry, whereas the tertiary sector handles forestry’s service activities, like eco-tourism and green financing. The percentage of secondary and tertiary industries rises with industrial upgrading and technical advancement, suggesting that forestry is progressively evolving into a service-oriented, technology-intensive sector and that the trend toward advanced industrial structure is becoming increasingly apparent. As a result, this ratio is a useful tool for assessing how well the forestry business is structured.

The industrial structure hierarchy coefficient method is used to measure the degree of rationalization of the forestry industry structure, according to Tang Zhan and Li Hongmei [48]. This is because the method can objectively reflect the coordination and rationality between the various levels within the industry by analyzing the proportionality of the various sectors of the industry. By determining the percentage of each industrial level, the industrial structure hierarchy coefficient approach assesses the level of structural optimization across various sectors, exposing the industrial structure’s rationality and growth potential. The following is the precise calculation formula:

$$F=\sum _{i=1}^n{W}_i{Q}_i\left(n=1,2,3\right)$$

(4)

The forestry industry structure’s degree of optimization is denoted by F; the weights of the forestry one, two, and three industries are indicated by W_i; the forestry one, two, and three industries are given the weights of 1, 2, and 3, respectively [11]; and the percentage of the forestry one, two, and three output values to the total forestry output value is indicated by Q_i.

3.4. Control Variables

The following control variables were used in this study to better regulate the variables that might have an impact on the overall factor productivity level of forestry ecology [44,49]: (1) Dependency on foreign trade: This study uses the ratio of total exports to GDP to describe the degree of reliance on foreign trade. A high degree of reliance on foreign trade may encourage the best use of resources, enhance the external benefits of forestry production, and consequently influence the ecological total factor productivity. (2) Degree of government intervention: The selection process is characterized by the ratio of government fiscal expenditure to GDP. Government intervention can encourage the implementation of forestry policies and the prudent use of resources, both of which have an impact on the expansion of ecological total factor productivity. (3) Development of the technology market: The technology market is characterized by the ratio of its turnover to GDP; it encourages technological innovation and application, improves ecological sustainability and forestry production efficiency, and influences ecological total factor productivity. (4) Level of innovation: the number of approved patents is a logarithmic indicator of technological and management innovation in forestry, which contributes to increased ecological efficiency and productivity. (5) Extent of natural disasters: The degree of natural disasters is measured as the ratio of the disaster-affected area to the disaster-affected area. Natural disasters harm forestry resources and diminish ecosystem service functions, which impacts the ecological total factor productivity. (6) Education spending: measured as the ratio of education spending to fiscal spending; higher education spending fosters the development of human capital, which advances technology and raises ecological total factor productivity in forestry. (7) Industrial structure: determined by the ratio of the primary industry’s output value to the overall output value; optimizing the agricultural industrial structure can improve the forestry ecology’s total factor productivity and make better use of land and resources. (8) Industrialization degree: determined by the ratio of increased industrial output value to regional GDP; higher industrialization might result in over-exploitation of resources, which can impact ecological sustainability and, consequently, forestry production efficiency. (9) Transportation infrastructure: The degree of transport infrastructure is determined by dividing the entire road area by the total population of the area at the end of the year. Better transport infrastructure can increase market accessibility and resource mobility, which in turn can impact forestry output. (10) Level of informatization: Determined by the number of Internet users in the area, informatization increases the efficiency of resource management, fosters total factor productivity in forestry ecology, and simplifies the flow of information and the use of technology. (11) Water infrastructure: defined as the ratio of the total area planted to crops to the effective irrigated area; it influences the use and distribution of water resources, enhances the environment for forestry production, and raises ecological total factor productivity.

3.5. Data Sources

In light of data availability, the data used in this study are based on panel data from 30 Chinese provinces from 2014 to 2022, primarily from the China Statistical Yearbook, the China Forestry and Grassland Statistical Yearbook of previous years, the EPS database, and the Wind database (given the limited availability or discontinuity of data in some regions; Tibet, Hong Kong, Macao, and Taiwan are excluded from this analysis). In order to fill in the gaps, this study linearly interpolates the province and city year data that are absent.

Table 2. Descriptive statistics.

Variable Name	Mean	Standard Deviation	Minimum Value	Maximum Values
Forestry ecological total factor productivity	1.0959	0.55951	0.0895	6.8898
Data elements	0.2074	0.4062	0	1
External trade dependence	0.2284	0.2458	0.0003	1.2156
Level of government intervention	0.2516	0.1018	0.1066	0.6430
Technology market development	0.0205	0.0319	0.0002	0.1910
Innovation level	9.8748	1.3200	6.4922	12.3990
Natural disaster	0.4786	0.2946	0	4.2063
Expenditure on education	0.1605	0.2596	0.1032	0.2166
Industrial structure	1.1310	0.6605	0.4944	5.2968
Industrialization	0.3087	0.0752	0.1008	0.4976
Level of Transportation facilities	11.7375	0.8525	9.4663	12.9126
Informatization level	0.0786	0.1596	0.0147	2.5129
Water infrastructure	0.4439	0.1787	0.1780	1.2325

provides comprehensive definitions, measurements, and statistical explanations of the variables.

3.6. Research Methodology

At present, most studies of policy effects assessment use traditional models such as the double difference model. However, there are problems such as model setting bias and linear assumption constraints, especially when the assumption preconditions are more difficult to satisfy for the parallel trend test, which results in biased estimation if used directly. At the same time, it is more difficult to overcome the problems of uncertainty in the form of confounding factor function, curse of dimensionality, regularization bias, and excessive attention to the “consistency” of traditional models. Dual machine learning can effectively reduce the errors caused by confounding factors and adapt to the diversified needs of practical scenarios by flexibly applying a variety of advanced algorithms to deal with complex data and non-linear relationships. At the same time, the method automatically corrects the hidden bias in the data through the two-stage computation framework, which improves the reliability of the results and the value of the practical application. In order to make up for the shortcomings of the traditional measurement model, this study adopts a dual machine learning model, whose basic regression model is a partially linear model [50]:

$$Y_{i,t+1}={\alpha }_0{\text{Event}}_{it}+g(X_{it})+U_{it}$$

(5)

$$E(U_{it}|Even{t}_{it},X_{it})=0$$

(6)

where: i and t denote city and year, respectively; Y_i,t+1 denotes the explanatory variable ecological total factor productivity in forestry; Event_it is the disposal variable, denoting the data element; α₀ is the disposal coefficient, with a conditional mean of 0; X_it is a set of high-dimensional control variables, which requires a machine learning algorithm to estimate the specific form $\widehat{g}$(X_it). In order to speed up convergence and ensure that the disposal coefficient estimates are still unbiased under small sample conditions, the following auxiliary regression model is constructed:

$${\text{Event}}_{it}=m(X_{it})+V_{it}$$

(7)

$$E(V_{it}|X_{it})=0$$

(8)

The procedure: First, a machine learning algorithm is used to estimate the auxiliary regression $\widehat{m}(X_{it})$ and take its residuals. Second, the machine learning algorithm is used to estimate $\widehat{g}$(X_it) to change the main regression form:

$$Y_{i,t+1}-\widehat{g}(X_{it})={\alpha }_{0}Even{t}_{it}+U_{it}$$

(9)

${\widehat{V}}_{it}$ is then regressed as an instrumental variable for Event_it to obtain unbiased coefficient estimates:

$${\alpha }_0={\left(\frac{1}{n}{\displaystyle \sum _{i\in I,t\in T}{\widehat{V}}_{it}}{\text{Event}}_{it}\right)}^{-1}\frac{1}{n}{\displaystyle \sum _{i\in I,t\in T}{\widehat{V}}_{it}}[Y_{i,t+1}-\widehat{g}(X_{it})]$$

(10)

Finally, the optimization process aims to accelerate the convergence of $\sqrt{n}\left({\alpha }_0-{\widehat{\alpha }}_0\right)$ to zero, a well-established principle in statistical modeling and numerical computation, ensuring more reliable and asymptotically unbiased estimates of the treatment coefficients. This convergence behavior is critical for achieving robust and efficient estimation in high-dimensional causal inference models, as emphasized in the literature on scientific computing and machine learning-based structural optimization.

4. Empirical Analysis

4.1. Base Regression Analysis

The effect of data elements on the total factor productivity of forestry ecology is empirically analyzed in this study using a dual machine learning model, which is based on the previous theoretical analysis and research design. The validation of inside and outside sample crossing, or “K-fold slicing”, is used to achieve the characteristics of improving the data utilization rate, avoiding overfitting, making the parameters more robust, etc. The best approach is the 5-fold crossing. This study uses the 5-fold crossover approach, which adopts a sample split ratio of 1:4, because it is the most effective. Neural network algorithms are also used to solve the primary regression and auxiliary regression.

Table 3. Benchmark regression.

Variable Name	(1)	(2)
	Forestry Ecological Total Factor Productivity	Forestry Ecological Total Factor Productivity
Data elements	0.3455 ***	0.5097 ***
	(0.1194)	(0.1755)
control variable with one term in the hierarchy	YES	YES
quadratic term of the control variable	NO	YES
time fixed effect	YES	YES
Province fixed effects	YES	YES
observed value	270	270

Note: *** denote significance at the 1% statistical levels, with robust standard errors in parentheses.

displays the results of the empirical regression.

As shown in Column (1) of Table 3, after including control variables, year fixed effects, and province fixed effects, the data element shows a significantly positive impact on forestry ecological TFP. To enhance model precision, Column (2) adds quadratic terms for the control variables. The results remain robust: the data element remains significant at the 1% level with a coefficient of 0.5097, indicating that each unit increase in the data factor corresponds to a 0.5097-unit increase in forestry ecological TFP.

4.2. Robustness Tests

4.2.1. Removal of Outlier Effects

This study uses the bilateral truncated-tail method to implement data cleaning for the entire sample in order to address the potential interference of extreme observations and the heterogeneity of regional-time-series distribution of forestry ecological total factor productivity under the SBM-GML measurement framework. All explanatory variables were adjusted for outliers at the 1% and 5% levels using the threshold setting, and the sample coverage was maximized to improve the validity of the estimation. The adjusted empirical analyses demonstrate that the absolute values of the coefficients are highly convergent with those of the base model estimates, and that the parameter estimates of the core explanatory variables, or the data elements, maintain positive characteristics at the 5% statistical significance level (results are detailed in column (1) of

Table 4. Robustness test table.

Variable Name	(1)		(2)		(3)
	Reduced Sample		Changing the Sample Split Ratio		Replacement Algorithm
	Shrinkage 1%	Shrinkage 5%	1:2	1:6	Mountain Ridge Return
Data elements	0.5372 ***	0.3762 **	3.1532 ***	0.3988 **	0.2614 ***
	(0.1723)	(0.1539)	(0.8382)	(0.1559)	(0.0994)
control variable with one term in the hierarchy	YES	YES	YES	YES	YES
quadratic term of the control variable	YES	YES	YES	YES	YES
time fixed effect	YES	YES	YES	YES	YES
Province fixed effects	YES	YES	YES	YES	YES
observed value	270	270	270	270	270

Note: ** and *** denote significance at the 5% and 1% statistical levels, respectively, with robust standard errors in parentheses.

). This demonstrates that the study’s policy insights are not consistently skewed by extreme data perturbations and that the original findings are robust to cross-sample dispersion.

4.2.2. Changing the Sample Split Ratio

Regression analyses were performed using sample split ratios of 1:2 and 1:6 to prevent discrepancies in results caused by sample split ratio settings, notwithstanding the claim that the 5-fold crossover approach—that is, a sample split ratio of 1:4—is ideal. The findings demonstrate (see Table 4, column (2)) that the significance of the impact of data elements on total factor productivity in forestry ecology does not change when the sample split ratio is reset. Additionally, the regression coefficient is positive, and the size does not differ from the baseline regression, which is adequate to demonstrate the validity of the initial findings.

4.2.3. Replacement Algorithm

To further investigate the potential impact of the algorithm on the findings of the baseline regression results, ridge regression was used in place of the original neural network algorithm in order to avoid the algorithm compromising the accuracy of the estimation results. The results demonstrate that the effect of data elements on forestry ecological total factor production remains significantly positive at the 1% level even when the algorithm is replaced with ridge regression (see Table 4, column (3)), and the baseline regression results remain valid.

4.3. Endogeneity Test

Considering that there are many factors affecting the total factor productivity level of forestry ecology, this study fails to completely control all relevant variables, so the endogeneity problem still exists. To address this issue, referring to Huang Qunhui et al.’s study [51], the interaction term between the number of post offices per million and Internet penetration in 1984 with one period of forward extrapolation was chosen as the instrumental variable. Meanwhile, drawing on Chernozhukov et al. [52], a partial linear instrumental variable model was constructed for analysis using a dual machine learning approach.

$$Y=0_{i}Event\times afte{r}_i+g(X)+u,e(UIX,DID)=0$$

(11)

$$IV=F\left(x\right)+V,E\left[VIX\right]=0$$

(12)

The relevance constraint and the exogeneity criterion must be met during the instrumental variable selection procedure. First, from a relevance standpoint, the Internet penetration rate lagging one period can represent the cumulative effect of technology diffusion, while the distribution density of post offices as a traditional information infrastructure may influence the design of later Internet infrastructure through path-dependence effects. The synergy between traditional and modern information dissemination facilities can be captured by the interaction term between the two, and there is a theoretical association with the current rate of Internet penetration. Furthermore, from the standpoint of exogeneity, the distribution of the number of post offices in 1984, as a historical data point, is primarily influenced by the distribution of administrative resources during the planned economy and is not directly related to the unobservable economic and social conditions of today. The exclusion criteria can be satisfied, and the interference of contemporaneous confounding factors further eliminated by creating an interaction term between this exogenous variable and the lagged Internet penetration rate.

By adding time-dynamic variables, this study creates a composite instrumental variable system with the goal of identifying structural discrepancies between the panel model and the cross-sectional data. The variable that inherits the exogenous benefit of historical instrumental variables and complies with the structural properties of panel data is specifically chosen to be the Internet penetration rate lagged by one period. After adjusting for endogeneity bias, the measurement results demonstrate that the positive promotion effect of data elements on forestry ecological total factor productivity is still statistically significant, and the estimated coefficients exhibit good robustness characteristics (see

Table 5. Endogeneity test table.

Variable Name	(1) IV: National Big Data Pilot Zone Lag Phase I	(2) IV: Post Office Counts in 1984 with Forward Extrapolated Period Internet Penetration Rates
Data elements	31.4022 *	2.5878 *
	(17.0719)	(1.5398)
control variable with one term in the hierarchy	YES	YES
quadratic term of the control variable	YES	YES
time fixed effect	YES	YES
Province fixed effects	YES	YES
observed value	270	270

Note: * denotes significance at the 10% statistical level, with robust standard errors in parentheses.

for specifics).

4.4. Conduction Mechanism Test

Data elements may be used in three effective ways to support the technological advancement of forestry, the sophisticated structure of the forestry industry, and the rationalization of the forestry industry’s structure to raise the level of ecological total factor productivity in forestry, according to this study’s analysis of existing literature and theories. However, the “two-step method” of Jiang Boat [53] attempts to test whether the core explanatory variables affect the explained variables through specific mechanism variables, taking into account the endogeneity bias of the mediation effect test, which states that the mediator variable and the outcome variable may be mutually causal. The technique uses two successive regression procedures to systematically examine the causal relationship between variables. The impact of data components on mechanism variables is confirmed in the first phase, and the impact of mechanism variables on total factor productivity in forestry ecology is explained in the second step using the literature.

Table 6. Mechanism tests.

Variable Name	(1)	(2)	(3)
	Technical Progress in Forestry	Rationalization of the Structure of the Forestry Industry	Advanced Forestry Industry Structure
Data elements	0.5156 **	0.5372 ***	0.3762 **
	(0.2519)	(0.1723)	(0.1539)
control variable with one term in the hierarchy	YES	YES	YES
quadratic term of the control variable	YES	YES	YES
time fixed effect	YES	YES	YES
Province fixed effects	YES	YES	YES
observed value	270	270	270

Note: ** and *** denote significant at the 5% and 1% statistical levels, respectively, with robust standard errors in parentheses.

displays the regression findings.

This study introduces forestry technological progress as a mechanism variable to be analyzed in order to test the mechanism of forestry technological progress in the data elements to enhance the total factor productivity of forestry ecology, based on the theoretical analysis in the previous section. In particular, a proxy variable for the advancement of forestry technology was the logarithm of the number of forestry patents. At the 5% level of statistical significance, the regression coefficient of forestry technological advancement is 0.5156, according to the regression results (See column 1 of Table 6). This paper further verifies that the effect of forestry technology progress on forestry ecological total factor productivity is significantly positive (See column 1 of

Table 7. Mechanism tests.

Variable Name	(1)	(2)	(3)
	Forestry Ecological Total Factor Productivity	Forestry Ecological Total Factor Productivity	Forestry Ecological Total Factor Productivity
Technical progress in forestry	0.2037 **
	(0.1024)
Rationalization of the structure of the forestry industry		0.2593 **
		(0.1051)
Advanced forestry industry structure			0.6011 ***
			(0.2341)
control variable with one term in the hierarchy	YES	YES	YES
quadratic term of the control variable	YES	YES	YES
time fixed effect	YES	YES	YES
Province fixed effects	YES	YES	YES
observed value	270	270	270

Note: ** and *** denote significance at the 5% and 1% statistical levels, respectively, with robust standard errors in parentheses.

). The study by Wang Huogen et al. [11] confirms that forestry technology progress can promote the improvement of forestry ecological total factor productivity, therefore, hypothesis H1 is confirmed. This result suggests that data elements can greatly promote forestry technology advancement. Data components, on the one hand, encourage cooperative innovation and the spread of forestry technologies. Particularly in isolated forest regions, the cross-regional and cross-organizational data sharing platform facilitates the quick spread of cutting-edge technologies and dismantles information barriers, hastening the adoption of green technologies like ecological restoration and water-saving farming. Concurrently, data integration has spurred cross-innovation between forestry and other domains, such as the integration of biotechnology and intelligent machinery, which has aided in the creation of ground-breaking technologies like carbon sink assessment and resilient breeding. However, by building an intelligent ecological monitoring and regulation system, technological advancements in forestry have greatly improved the ability of forestry ecosystems to endure disturbances and adjust to changes. This has improved ecological resilience, reduced the efficiency loss caused by ecological risks, optimized resource utilization and ecological service functions, and ultimately realized the sustained growth of the total factor productivity of forestry ecosystems.

This study presents forestry industry structure rationalization as a mechanism variable for analysis in order to examine the mechanism of forestry industry structure rationalization in data components to improve the total factor productivity of forestry ecology. In particular, the primary, secondary, and tertiary production values of forestry were assigned values using the industrial structure hierarchical coefficient approach as stand-in variables for the rationalization of the forestry industry structure. According to the regression results, the rationalization of the forestry sector structure has a regression coefficient of 0.5372, which is significant at the 1% statistical significance level (See column 2 of Table 6). According to this result, data elements can greatly aid in the rationalization of the forestry industry’s structure. This paper further verifies that optimization of forestry industry structure is significantly positive for forestry ecological total factor productivity (See column 2 of Table 7), and the study by Jin M et al. confirms that the rationalization of forestry industry structure can promote the improvement of forestry ecological total factor productivity [54], thus, hypothesis H2 is confirmed. On the one hand, the data element creates an industrial ecology that blends the specialized division of labor with the regional characteristic layout; on the other hand, it breaks the time and space limitations and information asymmetry of traditional forestry production; on the other hand, it realizes the intelligent matching of production factors through the construction of the industrial Internet platform; and on the other hand, it encourages each link of the industrial chain to shift from separate operation to synergistic development. However, by rearranging the production factors—land, labor, and capital—to focus on the production links with greater ecological benefits, the forestry industry’s structure is rationalized, which greatly increases the efficiency of factor combination. This optimization not only lessens the efficiency loss brought on by resource mismatch, but it also encourages the systematic improvement of the forestry ecology’s total factor productivity by making it easier to realize the value of ecological products.

Furthermore, this study presents the advanced forestry industry structure as a mechanism variable for analysis in order to examine the mechanism of the advanced forestry industry structure in improving the ecological total factor productivity of forestry via data components. In particular, a proxy variable for the advanced forestry industry structure was the ratio of the overall forestry output value to the sum of the output values of the secondary and tertiary forestry industries. According to the regression results, the advanced forestry industry structure regression coefficient is 0.3762, which is significant at the 5% level of statistical significance (See column 3 of Table 6). This paper further verifies that the advanced forestry industry is significantly positive to the total factor productivity of forestry ecology (See column 3 of Table 7), while the study of You et al. confirms that the advanced forestry industry structure can promote the improvement of the total factor productivity of forestry ecology [11], so the hypothesis H3 is confirmed. This result shows that data elements can significantly promote the advanced forestry industry structure. On the one hand, by enabling the digital transformation of forestry, encouraging the extension of the industrial value chain from traditional timber processing to knowledge-intensive service links, and encouraging the transformation of the advanced industrial structure, data elements give rise to new service industries like carbon sink measurement and ecological product value assessment. However, the advanced transformation of the forestry industry structure through technological innovation and service upgrading has greatly increased the industry’s capacity to create value, leading to higher economic benefits per unit of ecological resources. At the same time, this advanced transformation has pushed the industry to move toward low-carbon and green development, and through the application of cleaner production technologies and the valuation of eco-services, it has realized the benign mutual promotion of ecological protection and economic growth.

4.5. Heterogeneity Test

4.5.1. Green Finance

The impact of data elements in enhancing forestry eco-efficiency can be greatly impacted by variations in the degree of green finance development. The level of development of green finance, a market-oriented regulatory tool, significantly affects the availability of adequate funding and risk protection during the digital revolution of forestry. Innovative financial techniques like carbon sink trading and ecological compensation can be thoroughly linked with data components to create efficient market-oriented channels for the realization of forestry ecological value in areas where green finance is more ideal. This study uses the methodology of Li Sufeng et al. [55] to measure the level of green finance development in each province using seven secondary indicators, such as green fund, green support, green credit, etc., and to construct the interactions between the level of green finance development and the data elements of core construction variables for heterogeneity analysis. This is because variations in green finance may cause the role promoted by data elements to show different effects in different regions. The findings (Column 1 of

Table 8. Heterogeneity test.

Variable Name	(1) Green Finance	(2) Environmental Regulation	(3) Financial Autonomy
Data elements * Green finance	0.3621 *
	(0.1987)
Data elements * Environmental regulation		0.7424 **
		(0.3473)
Data element * Financial autonomy			0.8494 ***
			(0.3564)
Data elements	YES	YES	YES
green finance	YES	NO	NO
environmental regulation	NO	YES	NO
Financial autonomy	NO	NO	YES
control variable with one term in the hierarchy	YES	YES	YES
quadratic term of the control variable	YES	YES	YES
time fixed effect	YES	YES	YES
Province fixed effects	YES	YES	YES
observed value	270	270	270

Note: *, **, and *** denote significance at the 10%, 5%, and 1% statistical levels, respectively, with robust standard errors in parentheses.

) demonstrate that the interaction coefficient is positive and significant at the 10% level, suggesting that data elements have a bigger role in boosting total factor productivity in forestry ecology in places with higher levels of green finance. This outcome may be the result of more developed green finance areas having a more comprehensive mechanism for recognizing the value of ecological products and a market-oriented support system that can serve as a crucial assurance for the use of data elements. The financial risks and constraints of applying technology in the digital transformation of forestry have been successfully mitigated by financial tools like carbon trading platforms and green credits. At the same time, the transformation efficiency of ecological value driven by data elements has been greatly increased by relying on the ecological compensation mechanism, environmental rights and interests trading market, and other innovative institutional designs. Furthermore, a well-established green finance system fosters the growth of the market for forestry data services and new business models like ecological assessment and intelligent monitoring, which more completely converts the technical benefits of data elements into the improvement of forestry ecological total factor productivity. The market-based incentive mechanism based on technical empowerment is further strengthened by the synergy between data components and green finance, which serves as a twofold motivator for improving forestry eco-efficiency.

4.5.2. Intensity of Environmental Regulation

The binding and guiding effect of environmental regulations on forestry eco-efficiency varies greatly depending on their strength, and the stringency of regulatory standards has a direct impact on businesses’ investments in ecological protection, pollution prevention, and technological innovation. In general, areas with stricter environmental laws are better able to motivate forestry businesses to use cleaner production methods and support the shift of the forestry sector toward a low-carbon, greener future. The impact of data elements on the overall factor productivity of forestry ecology may therefore vary significantly by region due to variations in the level of environmental regulation. Therefore, based on Chen Shiyi et al. [56], the frequency of environment-related words in provincial government work reports from 2014 to 2022 was crawled as the degree of environmental regulation through text mining. The interaction between the intensity of environmental regulation and the data elements of core construction variables was constructed for the analysis of heterogeneity. This was done in order to investigate the impact of data elements on forestry ecological total factor productivity under different environmental regulations. The data elements contribute more strongly to forestry ecological total factor production in places with more environmental control, according to the results (Column 2 of Table 8), which also reveal that the interaction coefficient is positive and significant at the 5% level. The reason for this outcome could be that areas with stricter environmental laws have put in place better ecological protection frameworks and green development incentives, which can create a better policy climate for the use of data variables. In the meantime, depending on policy tools like eco-compensation and green certification greatly increases the impetus for the transformation of ecological value driven by data elements. Strict ecological protection standards and environmental regulatory measures encourage forestry enterprises to increase digital and intelligent technology investment to improve resource utilization efficiency. Strong environmental regulatory pressure also fosters the development of the forestry green technology innovation market and encourages the use of cutting-edge technologies like clean production and intelligent monitoring, which more successfully converts the technological benefits of data elements into real improvement of the total factor productivity of forestry ecology. In addition to strengthening legislative limitations and incentives based on technological empowerment, the synergistic effect of environmental legislation and data elements offers a double guarantee for the enhancement of forestry eco-efficiency.

4.5.3. Degree of Financial Autonomy

The degree of local financial autonomy directly affects their capacity to invest in the development of data infrastructure, the promotion and use of technology, and ecological compensation mechanisms. It also significantly influences the policy space and implementation strength of local governments in implementing forestry ecological construction and digital transformation. Therefore, the impact of data components on the total factor productivity of forestry ecology is anticipated to exhibit significant regional variation due to changes in the degree of financial autonomy. The GDP of each province in the current year was used to gauge economic development, and the relationship between financial autonomy and the data elements of core construction variables was built to examine the heterogeneity in order to gain a thorough understanding of the precise effects of data elements on forestry ecological total factor productivity under various economic conditions. According to the findings, the data elements have a bigger driving effect on forestry ecological total factor production in locations with a higher degree of financial autonomy (Table 8, column 3). The interaction coefficient is positive and significant at the 1% level. This outcome may be the result of more comprehensive forestry digitization support systems and more adaptable factor allocation mechanisms in regions with greater financial autonomy, which can serve as a crucial assurance for the release of data factor value. With the help of targeted local financial support, these regions can successfully cultivate forestry digitization service markets and new business subjects, enhancing the technology diffusion effect and resource allocation efficiency brought about by data elements. At the same time, by independently setting up financial special funds, these regions can concentrate on developing forestry big data platforms and intelligent monitoring facilities, lowering the technical threshold of data collection and application. Stronger financial autonomy can also assist local governments in creating unique ecological compensation systems, reducing market failure in the digital transformation of forestry through data-driven, precise subsidy policies, and more fully converting the technical benefits of data elements into meaningful improvements in the ecological efficiency of forestry. In addition to strengthening policy applicability and implementation effectiveness based on technological empowerment, the synergy between financial autonomy and data components offers institutional safeguards for the improvement of total factor productivity in forestry ecology.

5. Conclusions and Policy Recommendations

5.1. Conclusions

A dual machine learning model is chosen to empirically test the impact of data factors on the total factor productivity of forestry ecology and its mechanism. This study uses panel data from 30 Chinese provinces between 2014 and 2022. The SBM-GML model is used to measure the total factor productivity of forestry ecology in each province. A number of robustness and endogeneity tests support the empirical findings, which first demonstrate that the data factor significantly increases forestry ecological total factor productivity. Second, additional research shows that data factors can further boost forestry ecological total factor productivity, advance forestry technology, and rationalize and advance forestry industry structure. Third, the impact of data factors on forestry ecological total factor productivity is more pronounced in regions with high levels of financial autonomy, green finance, and environmental regulation.

5.2. Policy Recommendations

The aforementioned findings demonstrate that data components are a useful strategy for raising the forestry ecology’s total factor productivity. Consequently, policy recommendations are given in the following three categories based on the study’s findings.

The first step should be to expedite the production of data elements and to systematically encourage their in-depth application in the forestry area. To achieve the standardized gathering and connectivity of data from various sources, the development of a forestry big data platform that encompasses the entire range of forest resources, ecological monitoring, production, and operation should be prioritized. By enhancing data rights, pricing, and trading systems, a hybrid model of “government-led-market participation” can be developed for data sharing mechanisms that encourage the lawful circulation and release of the value of data pieces. Simultaneously, the system for protecting privacy and data security needs to be improved, particularly when hierarchical data management, including ecologically sensitive areas, is implemented. Regarding the implementation path, it is advised to choose regions with a stronger basis for digitizing forestry in order to conduct pilot demonstrations. These demonstrations should concentrate on innovations in Internet of Things sensing, remote sensing monitoring, and other critical technologies for data integration and application, as well as creating a model of experience that can be shared. In order to promote the long-term use of data components, colleges, universities, and research institutes should establish specialized training programs and enhance the development of forestry data talents.

Second, a system of data-driven innovation should be developed for the whole chain of forestry. In terms of technological advancement, it is advised to establish a national forestry digital technology innovation center that will concentrate on innovations in key technologies like digital twins and forest growth prediction models. Additionally, an innovation chain consisting of “data collection-algorithm research and development-application promotion” should be established. Simultaneously, the mechanism of industry-university-research collaboration should be enhanced, scientific research institutions’ data algorithms should be supported, and the docking of enterprise application scenarios should be deepened. In order to rationalize the industrial structure, a forestry big data platform should serve as the foundation for an industrial monitoring system that can detect factor mismatches between regions in real time and direct the best possible resource allocation through targeted policy intervention. It should specifically foster the composite development mode of “forestry +” and assist the data empowerment of new businesses like eco-tourism and forestry economy. Supporting the growth of professional service organizations like carbon sink measurement and ecological asset assessment, as well as enhancing the pertinent standard system and certification system, should be the policy focus for the advanced industrial structure. In order to establish a number of digital industry clusters in strategic forest regions and generate a positive interaction between technology, industry, and data, it is advised to undertake the “data elements +” industrial cluster cultivation program. This multifaceted, chain-wide policy design will completely unleash the data components’ catalytic effect on forestry upgrading and transformation.

Third, putting into practice a data factor empowerment plan that is tailored to a particular location. It is advised that areas spearheading the growth of green finance concentrate on creating an empowering model that combines data and money. In addition to enhancing data quality certification and valuation systems to lower financial transaction costs, pilot forestry carbon sink data asset trading platforms can be utilized to create financial products like carbon sink insurance and carbon sink pledges based on remotely sensed monitoring data. Data-driven environmental regulatory innovation should be the policy focus in areas with strict environmental regulations. Create a system for evaluating ecological performance based on ground monitoring and satellite remote sensing, and immediately connect the findings of data monitoring to policy instruments like ecological compensation and environmental taxes to establish targeted incentives. In order to gain experience for other regions, regions with a high level of financial autonomy are encouraged to investigate a variety of support policies for data elements, such as creating a subsidy system for data application and special bonds for forestry digitization. In order to encourage the spread of best practices through “twinning” and other channels, an interregional experience exchange framework must be established. The central financial administration can establish special transfer payments for regions that do not fit any of the three categories of characteristics, so that they can concentrate on helping to build their data infrastructure.

5.3. Limitations and Further Research

This study acknowledges several limitations. It relies on provincial-level panel data, which ensures broad spatial and temporal coverage but may overlook heterogeneity within regions. For example, variation among enterprises, forest plots, or households could be obscured at this aggregation level. We also approximate some variables (such as the data element indicator) using policy dummy proxies; while this approach aligns with previous research, it may not fully capture regional disparities or the evolving intensity of data utilization across provinces. Additionally, our dual machine learning framework effectively addresses high-dimensional confounding, yet it depends on the choice of algorithms and parameter tuning, and its “black box” nature can limit the transparency of the inferred causal mechanisms.

In light of these limitations, future research could pursue several directions. For example, incorporating more granular and multi-source data (e.g., firm-level surveys, household statistics, and satellite-based ecological measurements) could refine the measurement of forestry ecological TFP and uncover more nuanced mechanisms. Extending our analysis to other green sectors (such as grasslands and wetlands) or to international contexts could test the generalizability of our framework and explore how different institutional and contextual conditions influence data-driven improvements. Finally, further work could enhance model interpretability by using visualization tools (such as causal diagrams or Shapley value decomposition) and by developing region-specific policy simulations to more precisely evaluate how data elements facilitate the green transformation of forestry systems.