The Impact of Government Open Data on Firms’ Energy Efficiency: Analyse the Mediating Role of Capacity Utilization and Biased Technological Progress

Abstract

As a new type of production factor, releasing data dividends is of great significance in improving corporate energy efficiency. Based on the data of listed enterprises in China from 2011 to 2022, the establishment of government open data platforms in each prefecture-level city is taken as a policy shock event, and the impact of government open data on corporate energy efficiency is empirically examined through a multi-period DID model. The results show that government open data improves enterprise energy efficiency by approximately 2.5% (relative to the mean), and capacity utilization and biased technological progress are the main pathways of action. In addition, the application of big data technology can better fulfill the role of data factors in improving enterprise energy efficiency. Heterogeneity analysis finds that government open data has a stronger effect on enterprise energy efficiency improvement in areas with high manufacturing concentration, environmental tax rate leveling, and high Internet penetration. The study suggests that enterprises should apply big data technology and build a mechanism for integrating data assets and energy management so as to fulfill the important role of data elements in the green development of enterprises.

1. Introduction

The global race to harness data’s economic potential has accelerated dramatically in recent years. Developed economies were quick to recognize data’s strategic value. Data-driven productivity is essential for maintaining competitive advantage. Within the global landscape, China has developed a distinctive state-guided framework for data development. On 26 November 2019, the Central Committee for Comprehensively Deepening Reform formally designated data as a fifth factor of production alongside land, labor, capital, and technology through the opinions on building a market-based allocation system for factors. Subsequent policies further elevated its strategic role. The 2022 Opinions on Building a Data Base System emphasized leveraging China’s massive data scale and application scenarios to activate data potential, while the 2023 Overall Layout Plan for Digital China prioritized public data aggregation to unimpeded data resource circulation. Later, the Data Elements × Three-Year Action Plan (2024–2026) advocated integrating manufacturing and energy data—such as orders, scheduling, and power consumption—to enable applications like energy forecasting and gradient pricing. Collectively, these developments underscore that breaking data barriers and enhancing governance critically improve enterprise energy efficiency and economic quality.

Historically, during China’s rapid industrialization post-1978, reliance on natural resources, low-cost labor, and FDI fueled manufacturing expansion but encouraged crude energy use—particularly coal-intensive growth—resulting in high consumption and low efficiency. The community is faced with pressing ecological challenges linked to resource depletion and greenhouse gas emissions [1]. Since energy efficiency underpins economic-environmental sustainability, factor endowment theory suggests that initial factor conditions shape production modes, technological progress, and industrial structures, ultimately governing energy utilization. In the digital era, data emerges as a transformative factor; its replicability, near-zero marginal cost, and non-consumption optimize traditional factor allocation (e.g., capital and labor) by accelerating technology diffusion and refining resource flows, thereby elevating enterprise energy efficiency and offering new pathways for environmental progress.

This study’s threefold contributions include: (1) incorporating government open data (GOD) into energy efficiency analysis to systematize its impact mechanisms. (2) Empirically validating—via a two-stage strategy model—how GOD enhances energy efficiency through capacity utilization and biased technological progress. (3) Revealing heterogeneous effects of GOD under varying regional and industrial conditions, highlighting contextual barriers to data-driven efficiency gains. These insights provide critical theoretical and empirical support for advancing green development in China’s data-assetization era.

2. Policy Background and Theoretical Mechanism

2.1. Policy Background

China’s public data openness has evolved through four distinct phases, marked by progressive institutional refinement. During the germination period (2002–2007), policymakers prioritized information resource development, initiating e-government systems and foundational data infrastructure. The exploratory phase (2008–2014) shifted toward transparency, formalized by the Regulations on Government Information Disclosure, which established “disclosure as the norm, non-disclosure as the exception” as a core principle. A growth phase (2015–2018) accelerated public data sharing and openness, driven by the Big Data Development Action Outline. This period saw the creation of a unified national government data platform. Since 2019 (the development period), China has treated data as a strategic production factor, advancing authorized public data operations. Key milestones include the Opinions on Market-based Allocation of Factors and the establishment of the National Data Bureau to coordinate the data marketplace. This institutional groundwork directly enables initiatives like the “Data Elements×” Three-Year Action Plan (2024–2026), which incentivizes firm-level data innovation—a core focus of our study. By 2021, 24 provincial governments had launched open data platforms. Beijing’s platform supports smart-city development with transport/environment datasets; Shanghai’s exchange enhances market transparency via procurement/finance data; Guangdong offers 1000+ datasets (e.g., demographics, economics) to fuel AI/big data innovation.

2.2. Theoretical Mechanism

2.2.1. The Direct Impact of Government Public Data Opening on Enterprise Energy Efficiency

Data openness reduces information asymmetry [2], which can help enterprises carry out accurate energy use analysis and management optimization. Enterprises are able to optimize energy use and management by conducting big data analysis and intelligent decision-making through rich basic data resources. For example, the Zhejiang Provincial Energy Big Data Center in China integrates multi-source energy data from electricity, coal, petroleum, and other sources. The center has developed over 60 digital applications, including the Dual Carbon Smart Governance Platform and Energy Conservation and Carbon Reduction e-Account, to provide data support for government energy consumption regulation, corporate energy efficiency improvement, and social governance. The center effectively promotes corporate energy conservation, emissions reduction, and green transformation, providing data support and decision-making assistance for regional carbon peaking and carbon neutrality goals. Chen et al. (2024) [3] emphasizes the transformative potential of data-driven approaches in creating low-carbon, efficient solutions. Data openness also promotes technological innovation and R&D [4], especially cross-industry and cross-discipline collaborative innovation. On the one hand, more efficient and energy-saving technologies and equipment can promote the improvement of energy efficiency; on the other hand, collaborative innovation realizes the rational allocation of resources and effectively reduces energy waste. Lv et al. (2025) [5] point out that the aggregation of big data elements mainly aids firms in lowering energy consumption and emissions by increasing technical innovation and energy usage efficiency. In addition, enterprises can predict market demand and adjust production plans through data analysis, optimizing production capacity and energy factor input. Using the resource-saving effect of digital factors to improve the efficiency of using natural resources, especially energy, is vital for achieving a green recovery [6]. Therefore, government public data opening provides important support and guarantees for enterprises to improve energy efficiency. This study proposes Hypothesis 1.

Hypothesis 1.

Government public data opening helps to improve enterprise energy efficiency.

2.2.2. Mediating Mechanisms of Capacity Utilization

The capacity utilization rate reflects the efficiency of resource use and production capacity of enterprises. The open government public data platform provides rich industry information. Open government procurement data and industry statistics can help enterprises identify market opportunities and potential risks, which can help them make strategic adjustments and resource reallocation [7], thus adjusting production capacity more flexibly. At the same time, enterprises are able to predict market changes and demand fluctuations through big data, rationally arrange production plans, and optimize their capacity utilization. For example, Ma et al. (2020) [8] propose a big data-driven predictive production planning to improve energy and resource efficiency for energy-intensive manufacturing industries. In addition, data openness has enhanced the operational efficiency of enterprises [9]. Enterprises can work more closely with upstream and downstream supply chain partners to optimize supply chain management, reduce the transportation time and cost of raw materials and products, and improve overall capacity utilization. Efficient capacity utilization means that enterprises can make full use of their production resources, produce more products with the same energy consumption, and improve energy efficiency. By increasing capacity utilization, companies reduce their share of the fixed costs of energy. The energy consumption required for machines and equipment operating at low loads is spread over more products, increasing the energy efficiency per unit of product [10]. At the same time, optimizing capacity utilization can reduce energy wastage due to frequent adjustment of production lines and improve the overall efficiency of the production process and energy utilization. Based on this, this study proposes Hypothesis 2.

Hypothesis 2.

The government public data open platform improves enterprise energy efficiency by enhancing enterprise capacity utilization.

2.2.3. Mediating Mechanism of Biased Technological Progress

Technological progress is the fundamental way to improve the energy efficiency of enterprises in the government’s public data open platform. Kelley and Cheetham (1972) [11] pointed out that there are two kinds of bias in technological progress, capital and labor, which embody the role of technological progress in promoting the productivity of capital and the productivity of labor. With the development of environmental economics, energy-biased technological progress was gradually proposed. Public data enter the market as a highly efficient production factor, which can optimize the structure of the original production factor allocation and promote technological progress [12]. First, data resource inputs achieve capital-biased technological progress by promoting enterprise technology R&D and innovation [13]. Capital-biased technological progress can improve the efficiency of equipment and other infrastructure use and reduce energy waste. For example, automation and smart manufacturing technologies significantly reduce unnecessary energy consumption in the production process. Second, data resources promote labor-biased technological advances by empowering enterprise operations, alleviating information dissonance between links, reducing ‘friction’ between software and hardware, equipment and employees, and improving employee skills and productivity [14]. By improving labor productivity, enterprises reduce energy consumption per unit of output. Finally, enterprises use energy consumption data to identify inefficiencies in energy use, promote energy management innovation, and achieve energy-biased technological progress. Enterprises significantly reduce their energy consumption through high-efficiency energy equipment and energy-saving technologies, which directly improves energy use efficiency. Based on this, this study proposes Hypothesis 3.

Hypothesis 3.

The government public data open platform improves enterprise energy efficiency by promoting biased technological progress.

2.2.4. The Moderating Role of Big Data Application

Big data technology refers to a collection of technologies for collecting, storing, managing, and analyzing massive, diversified, fast-generating, and high-value data by means of advanced information technology, and its main features include large data volume, diverse data types, fast processing speed, and low value density. The Resource-Based View (RBV) argues that the competitive advantage of an organization stems from its unique resources and capabilities [15]. However, Luo and Chen (2024) [16] point out that having data resources is not enough to directly translate into competitive advantage for an enterprise, and the key lies in the ability of an enterprise to effectively integrate and allocate data resources and other strategic resources to unleash the true potential of data. Big data technology helps enterprises to acquire and apply information resources, which can better play the optimization role of data elements on enterprise energy management. Enterprises apply big data technology to deeply mine and analyze energy consumption data, policy information, and market trends to identify potential energy saving opportunities and efficiency improvement paths [17]. At the same time, big data technology helps enterprises to quickly process and analyze the government’s open public data to adjust energy use strategies and make more scientific decisions for energy efficiency improvement and resource allocation. The application of big data technology by enterprises can effectively enhance the role of the government’s open public data platform in improving the energy efficiency of enterprises. Accordingly, this paper proposes Hypothesis 4. The research framework of this paper is shown in Figure 1.

Hypothesis 4.

The application of big data technology can strengthen the role of government public data open platform on the enhancement of enterprise energy efficiency.

3. Research Design

3.1. Model Setting

3.1.1. Benchmark Regression Model

The establishment of a governmental public data open platform in each prefecture-level city is taken as a quasi-natural experiment to construct a multi-period double difference model. Among them, the explanatory variable ene_ijt denotes enterprise energy efficiency, the core explanatory variable pda_it denotes the establishment of a governmental public data open platform, X_ijt is the set of control variables, ${\gamma }_t$ and ${\mu }_i$ denote time and individual fixed effects, respectively, and ${\epsilon }_{it}$ denotes the random perturbation term.

$${ene}_{ijt}={\alpha }_1+{\beta }_1{pda}_{jt}+{\beta }_2{X}_{ijt}+{\mu }_i+{\gamma }_t+{\epsilon }_{ijt}$$

(1)

3.1.2. Two-Stage Strategy Model

In this paper, we refer to Liu et al. (2023) [18], who introduced a two-stage strategy model to test the transmission mechanism of public data openness on firms’ energy efficiency. Among them, M_ijt is the mediating variable, including capacity utilization and biased technological progress. In the first stage, this paper tests the impact of the establishment of a public data open platform on firms’ capacity utilization and biased technological progress through a double difference model. In the second stage, this paper uses the fitted values ($\widehat{M_{ijt}}$) of capacity utilization and biased technological progress estimated in the first stage to predict firms’ energy efficiency. The two-stage strategy model tests the transmission mechanism of the public data open platform on firms’ energy efficiency. The traditional three-step mediated effects model suffers from endogeneity in identifying the mechanism of action. The two-stage strategy model is similar to IV estimation, which not only emphasizes the stages of the main argument of the article but also decomposes the transmission channels among the variables.

$$M_{ijt}={\alpha }_3+{\rho }_1{pda}_{it}+{\rho }_2{X}_{ijt}+{\mu }_i+{\gamma }_t+{\epsilon }_{it}$$

(2)

$${ene}_{ijt}={\alpha }_4+{\phi }_1\widehat{M_{ijt}}+{\phi }_2{X}_{ijt}+{\mu }_i+{\gamma }_t+{\epsilon }_{it}$$

(3)

3.1.3. Moderating Effect Model

Enterprises can conduct data analysis more scientifically by applying big data technology, thus improving the ability to use data resources to manage enterprise energy efficiency. Therefore, this paper sets model (4) to test the moderating effect of big data application in the process of improving enterprise energy efficiency by public data open platform.

$${ene}_{ijt}={\alpha }_5+{\vartheta }_1{pda}_{jt}+{\vartheta }_2{bd}_{ijt}+{\vartheta }_3{pda}_{jt}{bd}_{ijt}+{\vartheta }_4{X}_{ijt}+{\mu }_i+{\gamma }_t+{\epsilon }_{ijt}$$

(4)

3.2. Data Sources

This paper takes Chinese A-share listed companies from 2011 to 2022 as the research sample and excludes sample companies with ST, PT, and sample companies with gearing ratios lower than 0 or higher than 1. Finally, 1381 companies in 189 prefecture-level cities and 78 industries are selected for the study. The data are mainly sourced from CSMAR, Wind, China Urban Statistical Yearbook, China Environmental Yearbook, and China Economic Information Network. The characteristics and distribution of the sample enterprises are shown in

Table 1. Distribution of Enterprises.

Feature	The Number of Enterprises	Proportion
The eastern region	930	0.673
Central and western regions	451	0.327
Private enterprise	660	0.478
State-owned enterprise	721	0.522
Number of employees ≤ 1000	212	0.154
1000 < Number of employees ≤ 5000	639	0.463
5000 < Number of employees	530	0.384

.

3.3. Variable Measurement

3.3.1. Explained Variables

Total factor energy efficiency (ene_ijt). The calculation of total factor energy efficiency needs to consider technical heterogeneity and capture the dynamic evolution of energy efficiency. This paper is based on the two-step stochastic frontier model proposed by Zhang and Tu (2022) [19] and Zhang and Zhou (2020) [20] for the base setting and improvement. First, it is assumed that the input factors in the production process of a firm include capital (K), labor (L), and energy (E), while the output is composed of desired output (Y) and undesired output (B). Among them, capital input is expressed as net fixed asset investment and deflated by the fixed asset investment price index; energy input is calculated by the coefficients of standard coal for each type of energy in the China Energy Statistical Yearbook; labor input is measured by the total number of employees in the enterprise; and desired output is expressed by the deflated sales revenue of the enterprise. For non-desired outputs, the pollution equivalents of major pollutants in industrial wastewater and exhaust gas are quantified based on the Measures for the Administration of Sewage Charge Collection Standards, and these pollution equivalents are summed up to comprehensively reflect the overall degree of pollution emission by the enterprise [21]. Second, based on the input-oriented Shepard’s energy distance, the ratio of potential energy inputs to real energy inputs is defined as total factor energy efficiency. When a firm’s energy inputs are located above the production possibility frontier surface, the ene_ijt value is 1, implying that the firm has achieved optimal efficiency in energy use. Conversely, below the frontier means that optimality has not yet been achieved, and the closer the ene_ijt value is to 1, the more energy efficient the firm is. Based on this, we calculated the total-factor energy efficiency indicator at the enterprise-year level. The calculation formula is as follows.

$${ene}_{ijt}={\displaystyle \frac{E/a}{E}}={\displaystyle \frac{1}{a}}={\displaystyle \frac{1}{D_E(K,L,E,Y,B)}}$$

(5)

In this paper, we further adopt the transcendental logarithmic production function and introduce a time trend term to capture the dynamic evolution of total factor energy efficiency so as to improve the measurement accuracy of the model, as shown in Equation (6). Where μ denotes the energy inefficiency value, which comes from the non-efficiency factors in the production process and is independent of the stochastic disturbance term ν.

$$\begin{gathered} -\mathit{ln}E=a_0+a_K\mathit{ln}K+a_L\mathit{ln}L+a_Y\mathit{ln}Y+a_B\mathit{ln}B+{\displaystyle \frac{1}{2}}{a}_{KK}\mathit{ln}K\times \mathit{ln}K \\ +a_{KL}\mathit{ln}K\times \mathit{ln}L+a_{KY}\mathit{ln}K\times \mathit{ln}Y \\ +a_{KB}\mathit{ln}K\times \mathit{ln}B+{\displaystyle \frac{1}{2}}{a}_{LL}\mathit{ln}L\times \mathit{ln}L+a_{LY}\mathit{ln}L\times \mathit{ln}Y+a_{LB}\mathit{ln}L\times \mathit{ln}B \\ +{\displaystyle \frac{1}{2}}{a}_{YY}\mathit{ln}Y\times \mathit{ln}Y+a_{YB}\mathit{ln}Y\times \mathit{ln}B \\ +{\displaystyle \frac{1}{2}}{a}_{BB}\mathit{ln}B\times \mathit{ln}B+a_{t}t+{\displaystyle \frac{1}{2}}{a}_{tt}\times t^2+a_{Kt}\times t\times \mathit{ln}K+a_{Lt}\times t\times \mathit{ln}L+ \\ {a}_{Yt}\times t\times \mathit{ln}Y+a_{Bt}\times t\times \mathit{ln}B+\nu -\mu \end{gathered}$$

(6)

In this paper, a bilateral stochastic frontier model is used for estimation. The first step is to estimate the within-group frontier surface. This paper follows Huang et al. (2015) [22] in estimating total factor productivity with consideration of technological heterogeneity and obtains the within-group energy efficiency estimate ${ene}_{ijt}^g=E(\mathit{exp}(-\mu ))$. Further, the error between the estimated and actual values of logarithmic energy inputs is obtained, i.e.:

$$\mathit{ln}{E}^{\mathrm{g}}(K,L,E,Y,B,\mathrm{t})=\mathit{ln}{\hat{E}}^{\mathrm{g}}(K,L,E,Y,B,\mathrm{t})+\tilde{\nu },\tilde{\nu }=\hat{v}-v$$

(7)

In the second step, the common frontier production technology is constructed and the gap between the in-group frontier and the common frontier is obtained $\ln TGD=\ln \left({\displaystyle \frac{D_E^m}{D_E^g}}\right)$. Combine the energy efficiency of the common frontier with that of the in-group frontier, i.e., ${ene}_{ijt}^m={\displaystyle \frac{1}{TGD}}\times {ene}_{ijt}^g$. Further, obtain the main regression equation, i.e.:

$$-\mathit{ln}{\widehat{E}}^{\mathrm{g}}(K,L,E,Y,B,t)=\mathit{ln}{E}^m(K,L,E,Y,B,\mathrm{t})+v^m-{\mu }^m$$

(8)

$v^m=-\tilde{v},{\mu }^m=\mathit{ln}TGD,TGD=E(\exp (-\mu ))$. Defining TGR as the inverse of TGD, the energy efficiency of the common frontier is expressed as:

$${ene}_{ijt}^m=TGR\times {ene}_{ijt}^g$$

(9)

3.3.2. Explanatory Variables

Government open data (pda_jt). The establishment of government public data open platforms in each prefecture-level city is taken as an exogenous policy shock. Specifically, pda_jt takes the value of 1 when a city has set up an open government public data platform in a certain year, and vice versa. Since 2012, China has successively established government public data opening platforms and built separate data opening websites such as the Shanghai Public Data Service Network, the Wuhan Public Data Opening Platform, and the Guangzhou Public Data Opening Platform. We refer to the relevant data from Fudan University’s “Report on the Opening of Public Data by Local Governments in China” to determine whether a city has established a government public data opening platform.

3.3.3. Mediating Variables

Capacity utilization (cu_ijt). Referring to Yang and Yan (2020) [23], the stochastic frontier analysis (SFA) method is used to measure the indicator, and the transcendental logarithmic production function is chosen to model the indicator, which is calculated as follows.

$$\begin{gathered} ln{Y}_{ijt}={\alpha }_0+{\alpha }_{1}t+{\displaystyle \frac{{\alpha }_2{t}^2}{2}}+{\alpha }_{3}ln{K}_{ijt}+{\alpha }_{4}ln{L}_{ijt}+{\alpha }_{5}t\times ln{K}_{ijt}+{\alpha }_{6}t\times ln{L}_{ijt}+ \\ {\displaystyle \frac{{\alpha }_{7}ln{K}_{ijt}\times ln{L}_{ijt}}{2}}+{\displaystyle \frac{{\alpha }_8{\left(ln{K}_{ijt}\right)}^2}{2}}+{\displaystyle \frac{{\alpha }_9{\left(ln{L}_{ijt}\right)}^2}{2}}+v_{ijt}-u_{ijt} \end{gathered}$$

(10)

$${cu}_{ijt}={\displaystyle \frac{E\left[f\left(x_{ijt},\beta \right)\exp \left(v_{ijt}-u_{ijt}\right)\right]}{E[f\left(x_{ijt},\beta \right)\left.\exp \left(v_{ijt}-u_{ijt}\right)\right|u_{ijt}=0]}}$$

(11)

In Equation (10), i stands for region, j stands for firm, and t stands for time. y stands for firm output measured in terms of business revenue; K stands for capital input measured in terms of fixed asset inputs; L stands for labor input measured in terms of the number of employees; v is a stochastic error term; and u is the technical loss error term used to compute technical inefficiencies.

Biased technical progress (ta_ijt). In assessing the contribution of factor inputs to output, Gong (2020) [24] suggests that factor output elasticity reflects not only the quality of input factors but also the effect of the application of technological innovation in economic production. Under the condition of constant input quantity, higher output elasticity can bring more input output, which is a direct reflection of factor-embedded technological innovation. Therefore, factor output elasticity can be regarded as an effective indicator of factor-embodied technological progress. This paper uses stochastic frontier analysis to estimate the output elasticities of firms’ capital, labor, and energy. As the transcendental logarithmic function can capture the interaction between different factors, this paper sets the model as a transcendental logarithmic function. The meanings of the relevant variables are the same as in the previous paper. We calculated capital-biased technological progress (tac_ijt), labor-biased technological progress (tal_ijt), and energy-biased technological progress (tae_ijt).

$$\begin{array}{l}\mathit{ln}{Y}_{it}=b_0+b_t\times t+{\displaystyle \frac{1}{2}}\times b_{tt}\times t^2+b_K\times \mathit{ln}{K}_{it}+b_L\times \mathit{ln}{L}_{it}+b_E\times \mathit{ln}{E}_{it}+{\displaystyle \frac{1}{2}}\times b_{KK}\times {\left(\mathit{ln}{K}_{it}\right)}^2\\ \hspace{1em}\hspace{1em}\hspace{1em}+{\displaystyle \frac{1}{2}}\times b_{LL}\times {\left(\mathit{ln}{L}_{it}\right)}^2+{\displaystyle \frac{1}{2}}\times b_{EE}\times {\left(\mathit{ln}{E}_{it}\right)}^2+b_{KL}\mathit{ln}{K}_{it}\times \mathit{ln}{L}_{it}+b_{KE}\mathit{ln}{K}_{it}\times \mathit{ln}{E}_{it}\\ \hspace{1em}\hspace{1em}\hspace{1em}+b_{LE}\mathit{ln}{L}_{it}\times \mathit{ln}{E}_{it}+b_{Kt}\times t\times \mathit{ln}{K}_{it}+b_{Lt}\times t\times \mathit{ln}{L}_{it}+b_{Et}\times t\times \mathit{ln}{E}_{it}+{\nu }_{it}-{\mu }_{it}\end{array}$$

(12)

3.3.4. Moderating Variable

Big data application (bd_ijt). This article measures the degree of big data application of a company by using the number of times keywords related to big data appear in the company’s annual report. We count the word frequency of keywords about big data, massive data, data centers, information assets, and arithmetic in the annual reports of listed enterprises. Finally, take the logarithm of the frequency after adding 1 as the indicator for enterprise big data application.

3.3.5. Control Variables

This paper selects the basic earnings per share (eper_ijt), capital intensity (cap_ijt), and net profit margin on total assets (ROA_ijt) at the enterprise level as control variables. Meanwhile, control variables at the regional level below the regional level are included. Economic scale (ln eco_it), expressed as the natural logarithm of the total per capita retail sales of goods in society, reflects the scale of regional economic development. The urbanization rate (urban_jt), measured by the ratio of the urban resident population to the total population, reflects the urbanization process of a region. The degree of openness to the outside world (open_jt) is measured by the ratio of a region’s total import and export volume to its GDP. The descriptive statistics of the variables refer to

Table 2. Descriptive Statistics.

VarName	N	Mean	SD	Min	Median	Max
eneijt	16,572	0.475	0.159	0.257	0.471	0.699
pdajt	16,572	0.409	0.492	0.000	0.000	1.000
eperijt	16,572	0.402	1.019	−10.170	0.260	49.930
capijt	16,572	2.623	4.182	0.089	1.857	289.885
ROAijt	16,572	0.036	0.062	−1.057	0.033	0.759
ecojt	16,572	2.971	1.318	0.496	2.843	6.878
urbanjt	16,572	0.670	0.131	0.350	0.666	0.896
openjt	16,572	0.493	0.370	0.000	0.465	1.548
cuijt	16,572	0.212	0.165	0.063	0.154	0.701
tacijt	16,572	0.160	0.028	0.038	0.158	0.272
taeijt	16,572	0.531	0.105	0.356	0.526	0.730
talijt	16,572	0.467	0.009	0.425	0.468	0.503
bdijt	16,572	0.555	0.931	0.000	0.000	5.303

. The histogram of the main variables is referred to in Figure 2.

4. Empirical Analysis

4.1. Parallel Trend Test

The Interactive Weighted Estimator (IW Estimator) proposed by Sun and Abraham (2021) [25] is particularly suitable for event studies. We use the IW Estimator to plot an event study diagram to determine whether parallel trends are satisfied. This requires that the development trends of the experimental group and the control group are consistent before the policy is implemented. This study selects the period immediately preceding the policy implementation as the baseline year to examine the differences between the experimental group and the control group over the four-year period before and after the policy implementation. Figure 3 shows the results of the parallel trend test. Meanwhile, we have provided coefficients, standard errors, and p-values in the Supplementary Materials. In Figure 3, prior to policy implementation, the confidence intervals for the explanatory variables largely encompassed the 0 point, indicating no significant differences between the experimental and control groups before the policy shock. Following policy implementation, the establishment of the data element open platform significantly enhanced the experimental group’s total factor energy efficiency. Therefore, the model satisfies the parallel trends assumption, enabling the difference-in-differences model to reasonably and effectively analyze the policy’s effects.

4.2. Benchmark Regression

Table 3. Benchmark estimation results.

Var		eneijt
	(1)	(2)	(3)
pdajt	0.004 **	0.004 **	0.012 ***
	(0.002)	(0.002)	(0.002)
eperijt		0.002 **	0.002 *
		(0.001)	(0.001)
capijt		0.000 **	0.000 *
		(0.000)	(0.000)
ROAijt		0.019 *	0.013
		(0.011)	(0.011)
ecojt			0.014 ***
			(0.001)
urbanjt			−0.506 ***
			(0.050)
openjt			0.131 ***
			(0.009)
time effect	YES	YES	YES
individual effect	YES	YES	YES
_cons	0.282 ***	0.280 ***	0.477 ***
	(0.001)	(0.001)	(0.025)
N	16,572	16,572	16,572

Note: The values in parentheses are clustered robust standard errors. * p < 0.1, ** p < 0.05, *** p < 0.01. pda_jt = government open data; ene_ijt = total factor energy efficiency; eper_ijt = earnings per share; cap_ijt = capital intensity; ROA_ijt = net profit margin on total assets; eco_jt = economic scale; urban_jt = urbanization rate; open_jt = openness to the outside world.

reports the estimation results of not including control variables, including firm-level control variables, and including district-level control variables. According to Table 3, we can clearly see that public data openness has a significant promotion effect on enterprise total factor energy efficiency, with an effect coefficient of 0.012, and passes the statistical test at the 1% significance level. This indicates that public data openness can effectively improve the enterprise’s energy efficiency, and this result verifies hypothesis 1.

Data openness reduces information asymmetry and enables enterprises to obtain higher-quality information so that they can make more accurate and efficient decisions in production and management. In addition, open data promotes technological innovation, and enterprises can use the data to develop more energy-efficient technologies and equipment, thereby improving energy efficiency. Data sharing also promotes cross-industry and cross-field collaborative innovation, enabling enterprises to achieve rational allocation of resources and synergies through shared data, optimize production processes, and further improve energy efficiency. At the same time, government open data supports enterprises in the application of big data analytics and intelligent decision-making, using advanced analytical tools and algorithms to optimize energy use and management. Together, these mechanisms make public data openness effective in improving the total factor energy efficiency of enterprises.

4.3. Robustness Test

4.3.1. Propensity Score Matching

In order to control for possible sample selection bias before treatment effects, this study uses the propensity score matching (PSM) method for robustness testing. In the first step, firm-level and district-level control variables are selected as covariates, and the control group sample that is most similar to the experimental group is found through caliper matching. The PSM method reduces the differences between the experimental and control groups. In the second step, DID regression was performed using the matched control group and the original experimental group. According to the results in column (1) of Table 3, public data openness still possesses a significant promotional effect on corporate energy efficiency, with an effect coefficient of 0.016, and passes the 1% significance level. This result further validates the effectiveness and robustness of governmental public data openness in promoting enterprise energy efficiency.

4.3.2. Excluding Municipalities Directly Under the Central Government

Municipalities directly under the central government have certain advantages in terms of resource endowment and policy support. In order to test the robustness of the findings, the sample data of enterprises located in the municipalities of Beijing, Shanghai, Tianjin, and Chongqing are excluded from the estimation again. After excluding these regions, the sample size is reduced from 1381 enterprises to 1108 enterprises. According to the estimation results in column (2) of Table 3, the estimated coefficient of public data openness is still significantly positive, with a specific value of 0.012, and passes the 1% significance level. Comparing the results with the baseline estimation, the sign and significance level of the coefficient of action remain unchanged, confirming the robustness of the findings of the previous study. This suggests that the establishment of the government’s open public data platform still has a facilitating effect on corporate energy efficiency, even if the sample of enterprises in municipalities with more favorable resource endowments and policy support is excluded.

4.3.3. Excluding the Impact of Other Policies

In 2018, the government issued the Pilot Work Program for the Construction of ‘Waste-Free Cities’, which aims to improve the quality of the urban environment by reducing the generation of solid waste and increasing the recycling of resources. The policy may have a potential impact on the energy efficiency of enterprises, leading to an overestimation of the coefficient value of the role of open government public data. In order to remove the interference of this policy effect, the sample of firms located in the pilot areas of ‘Waste-Free Cities’ is excluded and re-estimated in this paper. According to the results in column (3) of

Table 4. Robustness test.

Var	eneijt
	(1)	(2)	(3)	(4)
pdait	0.016 ***	0.012 ***	0.007 ***	0.013 ***
	(0.002)	(0.003)	(0.002)	(0.003)
control vars	YES	YES	YES	YES
time effect	YES	YES	YES	YES
individual effect	YES	YES	YES	YES
_cons	0.499	0.658 ***	0.678 ***	0.754 ***
	(0.024)	(0.045)	(0.040)	(0.049)
N	15,533	13,296	14,856	11,004

Note: The values in parentheses are clustered robust standard errors. *** represents p < 0.01. The first column is the estimated result after matching the propensity score; the second column is the estimated result excluding enterprises belonging to municipalities directly under the Central Government. The third column is the result of eliminating the policy effects of the “Pilot Work Plan for the Construction of a ‘Zero-Waste City’”. The fourth column is the result of eliminating the policy effect of energy consumption rights trading.

, the coefficient value of the government’s open public data platform on enterprise energy efficiency decreases to 0.007, which passes the significance level of one percent.

The energy use rights trading system regulates the energy use behavior of enterprises through the market mechanism, which may have a significant impact on enterprise energy efficiency. This paper further excludes the samples of enterprises in the pilot regions of energy rights trading to strip out the potential role of the energy rights trading system on the estimation results. The exclusion of these regions further reduces the sample size to 917. Based on the estimation results in column (4) of Table 4, the estimated coefficient value of pda_it is 0.013 and passes the test at the 1% significance level. This suggests that after stripping out other policy effects, the open government public data platform still significantly improves firms’ energy efficiency.

4.3.4. Tests for Other Explanatory Variables and IV Estimates

Data development quality. This article constructs an index system from five perspectives: data dissemination and sharing, data application environment, data element management, and data development and application to measure the quality of data development. We incorporated the new indicators into the model for regression estimation. The results show that the quality of data development still significantly improves the energy efficiency of enterprises by 1%, with a coefficient of effect of 0.573. This proves that the benchmark estimation results are robust.

To eliminate the endogenous influence of self-selection among cities, we selected the Internet penetration rate as the instrumental variable for 2SLS regression estimation. According to

Table 5. Other explanatory variables and IV estimates.

Var	eneijt	pdait	eneijt
	(1)	(2)	(3)
Scoreit	0.573 ***
	(0.025)
pdait			0.131 ***
			(0.015)
Internet penetration rate		0.011 ***
		(0.001)
F-test		212
control vars	YES	YES	YES
time effect	YES	YES	YES
individual effect	YES	YES	YES
N	16,572	16,572	16,572

Note: The values in parentheses are clustered robust standard errors. *** represents p < 0.01.

, the Internet penetration rate has prompted regions to actively establish public data opening platforms. The Internet can reflect the status of a city’s communication infrastructure. The government’s public data opening platform is associated with the urban communication infrastructure. Generally, the more developed the urban communication infrastructure is in a region, the more inclined it is to establish a public data opening platform. The second-stage results of the IV-2SLS estimation indicate that the establishment of the government’s public data open platform has indeed improved the energy efficiency of enterprises. The F-test is greater than 10. It indicates that the instrumental variable is valid. This proves the robustness of the research conclusion.

4.4. Mechanism Test

4.4.1. Capacity Utilization

Refer to

Table 6. Mechanism test results.

Var	First Stage				Second Stage
	(1) cuijt	(2) tacijt	(3) taeijt	(4) talijt	(5) eneijt	(6) eneijt	(7) eneijt	(8) eneijt
pdajt	0.001 **	0.001 ***	0.006 ***	0.000 ***
	(0.000)	(0.000)	(0.002)	(0.000)
w_cuijt					12.364 ***
					(1.778)
w_tacijt						10.760 ***
						(1.548)
w_taeijt							2.014 ***
							(0.290)
w_talijt								34.419 ***
								(4.950)
control vars	YES	YES	YES	YES	YES	YES	YES	YES
time effect	YES	YES	YES	YES	YES	YES	YES	YES
individual effect	YES	YES	YES	YES	YES	YES	YES	YES
_cons	0.232 ***	0.170 ***	0.511 ***	0.480 ***	−2.391 ***	−1.350 ***	−0.552 ***	−16.034 ***
	(0.001)	(0.009)	(0.027)	(0.002)	(0.410)	(0.261)	(0.147)	(2.371)
N	16,572	16,572	16,572	16,572	16,572	16,572	16,572	16,572

Note: The values in parentheses are clustered robust standard errors. ** p < 0.05, *** p < 0.01. pda_jt = government open data; ene_ijt = total factor energy efficiency; cu_ijt = capacity utilization; w_cu_ijt = the fitted value of the regression estimation of total factor energy efficiency by government open data; tac_ijt = capital-biased technological progress; tal_ijt = labor-biased technological progress; tae_ijt = energy-biased technological progress.

, the coefficient of the effect of public data openness on the capacity utilization of enterprises is 0.001, which is significant at the 5% level. The confidence intervals of the estimation results are referred to in

Table 7. The results of CI.

Var	Coefficient	CI
pdajt − tacijt	0.001	[0, 0.002]
pdajt − taeijt	0.006	[0.002, 0.009]
pdajt − talijt	0	[0, 0.001]
w_tacijt − eneijt	10.760	[7.724, 13.796]
w_taeijt − eneijt	2.014	[1.446, 2.582]
w_talijt − eneijt	34.419	[24.708, 44.129]

. Data analysis and optimization can help enterprises manage and utilize resources more effectively. Specifically, through data analytics, companies can identify energy inefficiencies or waste in the production process and take targeted measures to improve production processes and reduce energy consumption. In addition, data-driven decision-making and forecasting models can help companies schedule production more accurately and avoid excessive overcapacity.

In the second-stage estimation, the capacity utilization rate fitted by the first stage improves firms’ energy efficiency at the 1% significance level, with an effect coefficient of 12.364. When the capacity utilization rate increases, it means that firms make fuller use of existing production equipment, human resources, and other factors of production. At the same time, the fixed costs of production activities (e.g., depreciation of equipment, rent, overheads, etc.) can be spread over a larger number of products, which reduces the fixed costs per unit of product. This reduces idle and wasted resources, optimizes production processes, and results in lower energy consumption per unit of output, thus improving overall energy efficiency.

4.4.2. Biased Technological Progress

According to the estimation results of the first stage in Table 6, the government public data open platform has enhanced enterprise-biased technological progress at a significance level of 1%. Among them, the effect coefficient of energy-biased technological progress is the largest. This indicates that the establishment of the government’s public data open platform mainly promotes energy-biased technological progress in enterprises. The opening of government public data provides enterprises with abundant information resources, including energy usage data, environmental data, industry benchmark data, etc., which helps enterprises achieve scientific allocation of resources based on their own energy usage and the industry average level. Enterprises have effectively managed energy usage by improving information systems, enhancing energy-biased technological progress.

The fitted values of the first stage were incorporated into the second stage for regression estimation. The results indicated that enterprise-biased technological progress significantly improved the energy efficiency of enterprises. Among them, the effect coefficients of capital-biased technological progress and labor-oriented technological progress are higher. Li et al. (2023) [26] pointed out that the positive environmental effects of capital-biased technological progress have improved energy utilization efficiency. By leveraging information resources, enterprises can assess market supply and demand more accurately, adjust the allocation structure of capital and labor, optimize energy procurement and utilization plans, and thereby enhance energy utilization efficiency. Therefore, biased technological progress is an important way for government public data opening to improve the energy efficiency of enterprises.

5. Further Analyses

5.1. Moderating Effect of Big Data Application

According to the estimated results in column (1) of

Table 8. Test of the moderating effect of big data application.

Var	eneijt
	(1)	(2)
pdajt	0.011 ***	0.026 ***
	(0.002)	(0.002)
bdijt	0.008 ***
	(0.001)
pdajt bdijt		0.002 **
		(0.001)
control vars	YES	YES
time effect	YES	YES
individual effect	YES	YES
_cons	0.476 ***	−1.037 ***
	(0.025)	(0.011)
N	16,572	16,572

Note: The values in parentheses are clustered robust standard errors. ** p < 0.05, *** p < 0.01. pda_jt = government open data; ene_ijt = total factor energy efficiency; bd_ijt = big data application; pda_jt bd_ijt = the intersection item of government open data and big data application.

, big data technology improves the energy efficiency of enterprises at the 1% significance level, with an effect coefficient of 0.008. Big data technology combined with advanced technologies such as the Internet of Things, cloud computing, and other advanced technologies can build a smart energy management system. Enterprises monitor and control the use of energy in real time through the smart energy management system and discover and solve the failures and problems of the energy system in time. At the same time, it can also automatically adjust the strategy of energy supply and consumption according to the trend of energy data. Enterprises can use digital technology to optimize the energy supply network, improve the efficiency of energy transmission, and achieve efficient use of energy and energy saving and emission reduction [27].

Further, the interaction terms of big data application and explanatory variables are included in the model for regression. According to the estimated results in column (2) of Table 8, enterprises applying big data technology can obtain greater advantages from the public data open platform in the process of improving energy efficiency. The public data open platform provides abundant energy-related data, including energy consumption, energy supply, energy prices, and other aspects. Enterprises applying big data technology can efficiently collect, collate, and analyze these data so as to accurately grasp the actual situation and trend of energy use. In contrast, enterprises that do not apply big data technology may only rely on limited data sources and traditional data analysis methods, making it difficult to achieve a comprehensive and accurate picture of energy use. Through big data technologies, enterprises can deeply mine and analyze energy data to identify problems and potentials in energy use. For example, enterprises can analyze the energy consumption of different time periods and different equipment to find out the source of energy waste and the space for energy saving. In addition, big data technology can help enterprises predict future energy demand so as to formulate reasonable energy procurement and supply plans. All these functions greatly enhance the energy management efficiency of enterprises and reduce energy costs.

5.2. Heterogeneity Analysis

5.2.1. Manufacturing Agglomeration

After the regional manufacturing industry agglomerates and forms network clusters, it can improve the production efficiency of enterprises through the agglomeration effect. There may be heterogeneity in the impact of government public data opening on the energy efficiency of enterprises under different levels of manufacturing agglomeration. When the level of manufacturing agglomeration is high, enterprises usually rely more on public data to optimize production processes and improve energy efficiency. On the contrary, areas with a lower level of manufacturing agglomeration fail to maximize the effectiveness of data elements due to weaker agglomeration effects. In this paper, the sample is divided into high manufacturing agglomeration and low manufacturing agglomeration, and group regressions are conducted. Referring to Guo et al. (2020) [28], the level of manufacturing agglomeration is calculated using locational entropy with the following formula. In Equation (13), q_i is the number of employments in the manufacturing industry of region i; y_i is the total number of employments in all industries of region i; Q is the number of employments in the manufacturing industry of the whole country; Y is the total number of employments in all industries of the whole country. The sample is divided into high manufacturing agglomeration regions and low manufacturing agglomeration regions according to the average of the calculation results.

$$magg={\displaystyle \frac{q_i/y_i}{Q/Y}}$$

(13)

According to the grouped regression results in

Table 9. Results of heterogeneity test.

Var	eneijt
	Manufacturing Agglomeration	Manufacturing Non-Agglomeration	Increased Tax Burden	Unchanged Tax Burden	Developed Internet	Underdeveloped Internet
pdajt	0.033 ***	0.017 ***	0.019 ***	0.033 ***	0.008 ***	−0.006
	(0.003)	(0.002)	(0.002)	(0.003)	(0.002)	(0.006)
control vars	YES	YES	YES	YES	YES	YES
time effect	YES	YES	YES	YES	YES	YES
individual effect	YES	YES	YES	YES	YES	YES
_cons	0.499 ***	0.117 ***	−0.003	0.672 ***	0.592 ***	0.214 ***
	(0.047)	(0.030)	(0.042)	(0.039)	(0.036)	(0.073)
N	8061	8511	6840	9732	13,149	3423

Note: The values in parentheses are clustered robust standard errors. *** p < 0.01. pda_jt = government open data; ene_ijt = total factor energy efficiency.

, the establishment of public open data platforms promotes enterprise energy efficiency more in regions with high manufacturing agglomeration. Regions with high manufacturing agglomeration tend to be concentrated in the eastern coastal regions of China, such as the Pearl River Delta, Yangtze River Delta, and Bohai Rim. These regions have attracted a large number of manufacturing enterprises by virtue of their favorable geographical location, convenient transport conditions, and comprehensive industrial support. They have more policy support to incentivize technological advancement in energy management. Compared with regions with low manufacturing concentration, regions with high manufacturing concentration tend to have more concentrated resources and information, with rich industrial resources and enterprise networks. Public data openness can more effectively facilitate the dissemination and sharing of information in such environments. This makes it easier for firms to access valuable energy use data, which in turn incentivizes them to improve their own energy management. Enterprises in manufacturing agglomeration areas are often in complex industrial chains, and public data openness can lead to a smoother flow of information between upstream and downstream enterprises. Through collaboration between enterprises, resource management as well as power and energy supply can be optimized to improve the energy efficiency of enterprises in the region.

5.2.2. Environmental Regulation Effort

The introduction of the Environmental Protection Tax Law in 2018 marks the further improvement of China’s environmental protection legal system. The law transformed the traditional sewage charging system into a tax system, raised the legal status of environmental taxes and fees, and increased collection efforts. Since then, some regions have opted to raise environmental tax rates and enforce stricter environmental regulations. China is endeavoring to transition from a traditional economic model that relies on heavy and highly polluting industries to a greener, more sustainable one. Environmental regulations can motivate enterprises to reduce pollution emissions and promote a green transformation of the economic structure [29]. On the one hand, under the pressure of a high environmental tax burden, enterprises’ urgent need for energy efficiency improvement may prompt them to actively use public data resources to optimize energy management. On the other hand, the acquisition, processing, and analysis of public data require input costs and technical support, and an excessive tax burden may cause enterprises to cut back on production or squeeze the space for technical inputs to truly improve their energy efficiency. The samples of enterprises in the regions with tax burden uplift and tax burden leveling were subjected to regression analyses separately. According to the results in Table 9, GOD promotes the energy efficiency of enterprises in tax burden leveling regions more than in tax burden lifting regions. Although the Environmental Protection Tax Law is a policy that advocates green development, it does not work well with GOD to create a synergistic effect to promote enterprises’ energy efficiency. Tax burden leveling creates instability in environmental protection policies, which is not conducive to the formation of long-term investment expectations by enterprises. In tax burden leveling regions, a stable policy environment and less cost pressure provide resources and development space for enterprises’ long-term technology R&D and energy-saving investments to improve energy efficiency. On the premise of ensuring the realization of environmental protection objectives, the government should reasonably adjust the level of tax burden, improve policy support and incentive mechanisms, and encourage enterprises to invest in environmental protection technology upgrades and energy efficiency improvements.

5.2.3. Information Technology

The government’s public data opening platform establishes a unified national system by bridging communication between data providers and users. This enables the concentration, orderly opening, and standardized use of public data, fostering innovation-driven growth in new industries and business models. However, enterprises’ ability to utilize these resources hinges critically on regional digital infrastructure and data application capabilities. Our analysis reveals a stark regional disparity. In regions with developed internet (Table 9), OGD significantly improves energy efficiency. Firms leverage internet access to acquire public data, enabling advanced analytics for energy optimization, cost reduction, and collaborative innovation across supply chains. In regions with underdeveloped internet (Table 9), OGD shows no statistically significant effect. This underscores that digital capacity is a prerequisite for OGD absorption. Without foundational infrastructure and complementary technologies, firms lack the capability to interpret, analyze, and operationalize open data. It limits utility for energy management. Thus, while IT-developed regions achieve win-win economic-environmental gains through OGD-integrated cleaner production, regions lacking digital readiness fail to convert data access into efficiency improvements. This highlights the necessity of parallel investments in digital infrastructure to unlock OGD’s potential universally.

6. Conclusions and Policy Recommendation

Based on the sample data of Chinese listed enterprises during 2011–2022, this paper empirically examines the impact of the government public data open platform on enterprise energy efficiency through a multi-period DID model using the establishment of the government public data open platform in each prefecture-level city as a policy shock event. The findings show that the establishment of the government public data open platform can improve enterprise energy efficiency, and the effect is enhanced over time cumulatively. The findings pass a series of endogeneity and robustness tests. The results of the mechanism test indicate that capacity utilization improvement and energy-biased technological progress are the main ways in which GOD improves enterprises’ energy efficiency. In addition, the application of big data technology facilitates enterprises to fully explore, analyze, and utilize data resources, which enhances the positive effect of GOD on enterprise energy efficiency. Heterogeneity analysis reveals that in regions with a high degree of manufacturing agglomeration, environmental tax rate leveling, and high IT penetration, open government public data has a greater effect on the improvement of corporate energy efficiency. This provides a validated policy blueprint for emerging economies seeking industrial decarbonization, showing that replicating such platforms—especially in manufacturing hubs—could accelerate global climate goals by bridging critical implementation gaps like the digital divide. The findings further inform international standards and trade policies (e.g., carbon border adjustments), while validating that OECD nations could achieve industrial energy savings through similar data-driven approaches.

Based on the findings, this paper proposes the following policy recommendations. First, establish secure data-exchange ecosystems integrating enterprise energy management with public data assets. This enables industry-wide circulation of validated datasets, empowering firms to deploy digital tools for granular energy waste identification and dynamic allocation. Second, leverage these technological foundations—scalable beyond sampled firms through targeted subsidies—to institutionalize data- and AI-enabled efficiency upgrades. Smart monitoring networks and adaptive learning systems can synchronize energy supply with demand fluctuations, transforming firm-level efficiency gains into regional resource savings where coordinated adoption occurs. Third, strategically guide higher-stage manufacturing agglomeration through specialized industrial zoning and shared infrastructure. Such clustering amplifies knowledge spillovers and data-driven efficiency multipliers—enabling regional scalability when integrated with the above digital frameworks. These enterprise-level interventions achieve aggregate impact though technology adoption is accelerated via green subsidies, data exchange protocols respect operational boundaries, and agglomeration policies leverage localized spillover potential.

The study acknowledges the following limitations. Our sample is restricted to listed firms. These companies typically possess greater resources, standardized management, and policy responsiveness compared to SMEs and non-listed enterprises. This limits the generalizability of our findings, as SMEs may experience divergent effects from public data openness policies due to more severe constraints in data access, processing, and application capabilities. Expanding the research object and identifying the micro effects of policies will be a valuable research direction in the future, providing more solid and comprehensive scientific support for theoretical development and policy formulation. Future research can be further deepened by the following: (1) expanding analysis to SMEs through enterprise-level data collection or surveys to study policy impacts, identify unique challenges (e.g., technical capacity gaps), and evaluate needs for differentiated policy support; (2) developing simulation approaches to estimate policy effects on SMEs where granular data remains scarce; and (3) combining our micro-foundations with regional input-output tables or integrated assessment models could yield robust aggregate projections.