Green Value from Technology Finance Policies Towards Sustainability: Evidence of a Quasi-Natural Experiment on Urban Carbon Reduction in China

Abstract

Innovation can balance environmental objectives while enhancing economic efficiency. However, its policy dimension has largely been overlooked. Leveraging financial instruments to promote technological innovation has become a primary policy tool. This study treats China’s pilot policies on technology finance integration as a quasi-natural experiment and constructs a multi-period difference-in-differences model with 5434 panel datapoints from 286 cities between 2005 and 2023. Within the environmental Kuznets curve theoretical framework, we examine the impact of technology finance policies on urban carbon emissions. The findings reveal that these policies significantly curb urban carbon emissions. Heterogeneity analysis shows stronger effects in large, non-resource-dependent eastern cities than in others. Mechanism tests indicate that these policies primarily generate green value through technological innovation and industrial structure upgrading, with the latter demonstrating a significantly stronger effect than the former. Our findings regarding the outstanding role of industrial structure upgrading may challenge and reorient the literature’s general focus on green technologies. This study employs economic principles to offer new insights into innovation’s effective balancing of economic growth and sustainability, broadening discussions about the green value of financial policy tools.

1. Introduction

With global challenges such as climate change, biodiversity loss, and resource depletion intensifying, governments and the public increasingly urge enterprises to address environmental issues. However, this means enterprises face more resource constraints and higher environmental costs, hindering their core goal of profit maximization. Such divergence between environmental and business goals has led to unsustainable or even fraudulent environmental practices, such as greenwashing [1]. A growing body of literature argues that to encourage businesses to engage in sustainable, long-term, and voluntary environmental behavior, it is essential to explore the dimension of interests, which requires achieving environmental objectives while maintaining economic efficiency [2,3]. Government policy constitutes a credible vehicle for simultaneously attaining both objectives. Therefore, this study focuses on China’s financial innovation policies, implemented at the urban level, and employs a difference-in-differences (DID) design to generate novel empirical evidence.

As the core driving force of contemporary social development, technological innovation is profoundly changing all aspects of the economy, society, and human life [4]. In particular, technological innovation provides feasible solutions to ease the tension between economic growth and green low-carbon development, emerging as the latest frontier in sustainability research. In major economies, technological progress contributes over 60% to broad economic growth and nearly 30% to industrial carbon reduction [5].

Given the current perspective on technological innovation and its financial support system, three gaps strongly spark our research interest. First, the literature exhibits conflicting conclusions. While research on how technological innovation affects sustainability is active, the conclusions remain divergent. Some scholars contend that advancements in technological innovation reduce per capita carbon emissions: the adoption of new technologies enhances resource-use efficiency, reduces reliance on traditional energy, and generates substantial green dividends [6,7]. By contrast, other empirical studies reveal that technological innovation drives regional economic growth, thereby expanding the scale of carbon emissions [8,9]. The invention and application of new technologies have significantly boosted industrial production scale and efficiency, exacerbating carbon emissions [10]. These conflicting hypotheses require further empirical verification.

Second, the green value of technology finance (tech-finance) is still unclear [11]. The literature predominantly focuses on green finance mechanisms while overlooking the rapidly evolving domain of tech-finance. Serving as the capital nexus throughout the innovation lifecycle, tech-finance plays a critical role in capital allocation, risk diversification, resource integration, and industrial transformation. It is widely implemented across major economies, including the U.S., China, the EU, Japan, and the U.K. [12]. As innovation-driven growth becomes a global strategic priority, leveraging financial tools to spur technological innovation has emerged as a core policy agenda for emerging economies [13]. While academic significantly lags behind the rapid practical advancements in tech-finance [14].

Third, studies from policy-driven perspectives remain insufficient. Prior studies predominantly use financial institutions’ technology-related credit balances as proxies, focusing on capital quantity while neglecting tech-finance’s inherent policy-driven nature [15]. This sector faces systemic challenges: high innovation failure rates, absence of reliable collateral, extended capital recovery periods, and adverse selection and moral hazard induced by information asymmetry [16,17,18]. These factors expose lenders to significant credit and liquidity risks, as evidenced by the 2023 collapse of Silicon Valley Bank. Consequently, financial institutions lack endogenous motivation for technology lending, necessitating external policy drivers, particularly government intervention [19]. Unlike conventional technology credit, policy-driven tech-finance emphasizes resource guidance over profitability, integrating innovation, finance, and governance for multi-dimensional impacts [20]. Current conclusions about tech-finance’s environmental effects may diverge from policy outcomes, as analyzing mere financial intensity overlooks systemic policy interactions. A policy-focused approach, however, enables a rigorous evaluation of financial policy effectiveness, thereby enriching theoretical frameworks on finance and sustainability linkages.

The aforementioned unresolved and under-explored areas provide new research space for this study. We draw on the panel data of 286 prefecture-level cities in mainland China from 2005 to 2023, taking the “Pilot Program for Promoting Integration of Technology and Finance” implementation plans in 2011 and 2016 as policy shocks. Thus, we conduct a quasi-natural experiment using a DID model. This study aims to systematically test and analyze the “green” effectiveness of tech-finance policies, offering new approaches to balancing high-quality economic growth and green low-carbon development and enriching the research system on financial mechanisms and policy orientations. The core research questions are as follows: clarifying the direction and intensity of tech-finance policies’ impact on carbon reduction; identifying specific mechanisms through technological innovation and industrial structure upgrading; quantitatively analyzing heterogeneity across cities in terms of geographical location, economic scale, and resource endowments.

The findings reveal that tech-finance policies significantly curb carbon emissions, with this inhibitory effect varying across regions with different geographical locations, economic scales, and resource endowments. Additionally, tech-finance policies reduce carbon emissions mainly by boosting technological innovation and industrial structure upgrading, with the latter playing a significantly stronger role than the former. Using economic principles, this study provides new insights into how innovation effectively balances economic growth and sustainability, with the following marginal contributions: (1) Unlike previous studies focusing on tech-credit volume indicators, it centers on the policy dimension of tech-finance, deepening our understanding of how financial policies drive carbon reduction mechanisms. (2) It conducts heterogeneity analysis based on geographical location, economic scale, and resource endowments, identifying variations in tech-finance policies’ carbon reduction effects across city types and providing empirical evidence for region-specific policymaking. (3) It tests and explains the theoretical mechanism by which tech-finance policies curb carbon emissions through the synergy of technological and structural effects, enriching the theoretical framework linking financial policies to low-carbon development. (4) Our findings on the prominent role of industrial structure upgrading may challenge and reorient the literature’s general focus on new green technologies and broaden discussions on the green value of financial policy tools.

The content of this study is divided into four sections. Section 2 introduces the selection of the research object, along with the mechanism analysis and research hypotheses based on the environmental Kuznets curve theory. Section 3 elaborates on the DID model, variable selection, research samples, and data statistics. The empirical results of the DID model, as well as the results of heterogeneity and mechanism analysis, are all detailed in Section 4. That section also includes the corresponding results of endogeneity tests and robustness tests. Section 5 discusses the conclusions, implications, contributions, and limitations of this study and proposes prospects for future research.

2. Theoretical Analysis and Hypotheses

Based on the literature review presented earlier, three key points can be identified: (i) the direction of the impact of innovation on carbon emissions remains unclear; (ii) while the green value of green finance has received widespread attention, that of tech-finance has not been sufficiently discussed; (iii) theoretically, policy-driven finance should have environmental benefits, yet empirical research to verify them remains scarce. These provide us with dimensions for selecting research objects.

The public externality of policies, rather than the profit motive, provides the logical possibility for tech-finance to achieve dual objectives. Therefore, this study focuses on tech-finance policies instead of tech-finance itself.

Table 1. Selection of research subjects.

Dimensions of Selection	Tech-Finance Policies	Tech-Finance	Comparing and Selecting Results
Purpose attribute	Public externalities	Profit-driven	Balance of dual objectives: Tech-finance policies
Direction of technological innovation	Social welfare	Corporate interest	Environmentally friendly possibilities: Tech-finance policies
Current research attention	Less	More	Possible research frontiers: Tech-finance policies

presents the comparative selection process and results for the two research objects.

2.1. Background of Technology Finance Policies

In December 2010, China’s Ministry of Science and Technology, the central bank of China, and three other ministries jointly promulgated the “Pilot Implementation Plan for Promoting Technology–Finance Integration”. Launched initially in 2011, the program designated 41 cities across 16 regions as the first cohort of pilot areas. Its core objectives included alleviating financing constraints for technology enterprises and accelerating the commercialization of scientific and technological achievements [21]. Subsequently, building on the pilot’s preliminary success, the five ministries expanded the project’s coverage in 2016 by incorporating nine additional cities into the second batch of pilot regions. Through systematic integration of venture capital, credit financing, and capital market resources, the initiative enhanced financial institutions’ capacity to support technological innovation, reduced financing barriers for corporate R&D activities, and fostered deep synergy between technological advancement and financial capital allocation. The two pilot policies were implemented in 2011 and 2016, providing relatively exogenous, quasi-natural experimental conditions for this study [22].

Drawing on the established literature [23,24], this study adopts carbon emissions as a proxy for sustainability. All pilot regions examined herein for implementing tech-finance policies are urban areas. Given the role of cities as primary regional hubs for economic activities and production factor concentration, their CO₂ emissions constitute over 70% of China’s total [25], rendering them highly representative for research purposes. Consequently, subsequent analyses of environmental protection practices in this study focus exclusively on urban carbon emissions.

2.2. Direct Impact of Tech-Finance Policies on Urban Carbon Emissions

Unlike a financial tool, which pursues high returns via high-risk investments, tech-finance policies center on “policy”, specifically, government guidance on the allocation of financial resources. Consequently, tech-finance policies exhibit distinct publicness and externalities. For example, in 2024, the central bank of China launched a special credit support program, setting aside CNY 500 billion in a tech-finance fund. Of this fund, CNY 100 billion was earmarked as a special quota, targeting tech-based small and micro-enterprises, focusing on resolving their initial financing difficulties [26]. This special program offers highly preferential interest rates. Its aim is not profit but rather to build a financial support system that encourages financial institutions to strengthen their engagement in tech-finance and enhance financing support for tech-based small- and medium-sized enterprises (SMEs) [27].

Compared with tech-finance itself, tech-finance policies incorporate strong policy orientations and public externalities. Under government policy guidance, tech-finance slows its profit-driven momentum, instead prioritizing tech-based and innovative enterprises as key support targets. This has fostered strategic emerging industries with core technologies and numerous low-carbon industries [28]. Through fiscal incentives and financial empowerment, pilot policies have strengthened R&D support in pilot cities, alleviated corporate financing constraints, and directed capital toward enterprises with higher environmental awareness, supporting urban economic development and industrial transformation under the “green” goals [29]. This not only drives the development of low-carbon industries and helps pilot cities explore new development models, but also promotes industrial structure upgrading and optimization, guiding industrial low-carbon transformation [30]. Pilot cities utilize new energy technology R&D and applications to improve resource utilization efficiency, achieving sustained carbon emission reduction. Additionally, enterprises seize development opportunities and pursue low-financing-cost projects, driving the development of low-carbon industries, advancing pilot cities’ exploration of new development models, and effectively facilitating industrial structure upgrading and rationalization [31]. Therefore, we propose the following hypothesis:

Hypothesis 1. 

Tech-finance policies have an inhibitory effect on urban carbon emissions.

2.3. Transmission Mechanism of Tech-Finance Policies in Urban Carbon Emissions

The environmental Kuznets curve (EKC) theory posits an inverted U-shaped relationship between economic growth and environmental degradation intensity [32]. Initially, environmental degradation intensifies with economic growth, but after reaching a critical development threshold, pollution levels begin to decline, and environmental quality gradually improves [33]. According to scholars’ latest theoretical interpretations of the EKC, three pathways through which economic development influences environmental quality have been proposed, namely, scale effects, technological effects, and structural effects [34,35].

(i) Scale effects: Economic growth negatively impacts environmental quality in two ways. Firstly, economic growth entails increased inputs, which boosts resource consumption [36,37]. Secondly, higher output leads to greater pollution emissions [38]. (ii) Technological effects: Higher income levels correlate with improved environmental protection technologies and enhanced efficiency. As economies grow, rising R&D expenditures drive technological progress with dual impacts. Productivity gains improve resource utilization efficiency, reducing factor input per unit of output and mitigating production impacts on ecosystems [39]. In addition, clean technologies replace polluting technologies, enabling resource recycling and reducing pollution emissions per unit of output [40]. (iii) Structural effects: Income growth induces transformations in both output and input structures. Early-stage shifts from agriculture to energy-intensive heavy industries increase pollution, while subsequent transitions to low-pollution service sectors and knowledge-intensive industries reduce emission intensity through optimized input–output configurations [41].

While scale effects exacerbate environmental degradation, technological and structural effects generate remedial impacts. At advanced development stages, the latter two effects dominate, slowing pollution accumulation. Social production enhances environmental quality through two channels: technological progress improves resource efficiency and reduces production impacts on nature, while industrial restructuring achieves sustainable emission reductions by upgrading regional industrial structures. These mechanisms collectively promote green and low-carbon development. Figure 1 shows the EKC theoretical framework and its triple-effect mechanism. Among the three, technological effects and structural effects form the theoretical basis for this study’s explanation of the transmission mechanism. Based on the above analysis of the EKC theory, this study proposes the following two hypotheses:

Hypothesis 2. 

Tech-finance policies reduce urban carbon emissions by promoting technological innovation.

Hypothesis 3. 

Tech-finance policies reduce urban carbon emissions by driving industrial structure upgrading.

3. Materials and Methods

3.1. Models

The cornerstone of rigorously evaluating the impact of tech-finance policies on urban carbon emissions lies in establishing a causal nexus between them. The DID model, recognized as a robust methodological framework for quasi-natural experiments, provides an ideal analytical tool for this purpose [42]. The primary frontier area lies in the cohort-specific estimation of staggered DID and the integration of DID with the event study methodology. The purpose is to characterize the time path of policy effects and address the challenge of “negative weighting” arising from heterogeneous treatment timing [43]. Against this backdrop, we employ a staggered DID specification to empirically identify the causal relationship between tech-finance policies and urban carbon emissions. The specific DID model is presented in Equation (1):

$$C{e}_{it}={\alpha }_1+{\beta }_{1}Di{d}_{it}+{\displaystyle \sum {\theta }_n}c{v}_{nit}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(1)

where i represents the city, t denotes the year, and n is the number of control variables. Ce_it is the dependent variable, representing the carbon emission levels of the i-city in the t-year; Did_it is the policy variable of this study and the core explanatory variable, with its coefficient, β, being the key parameter of interest, reflecting the net effect of tech-finance policies on urban carbon emissions; cv represents a set of control variables, with n being the number of control variables; θ is the corresponding coefficient for each control variable; μ represents the individual fixed effect; λ is the time-fixed effect; and ξ represents the random disturbance term.

3.2. Variables

3.2.1. Dependent Variable

Proxy variables for measuring carbon emissions at the city level include total or incremental CO₂ emissions, CO₂ intensity (emissions per unit GDP), and per capita CO₂ emissions. As concentrated hubs of population and economic activities, urban carbon emissions inherently arise from the interplay of industrial production and residential consumption. To ensure cross-sectional comparability and mitigate confounding effects from population-scale heterogeneity across cities, this study employs the natural logarithm of per capita CO₂ emissions as the core metric for urban carbon performance [44,45].

3.2.2. Independent Variable

The DID method is a rigorous quasi-experimental framework for policy evaluation, with its core explanatory variable constructed as the product of a treatment dummy (Treat) and a time dummy (Time) [46,47]. Pursuant to the “Pilot Implementation Plan for Promoting Tech–Finance Integration” promulgated by the Ministry of Science and Technology and four other departments, two policy waves were executed: the first in 2011 designating 41 pilot cities and the second in 2016 adding 9 new regions. Notably, Ningbo was double-counted in the initial cohort, resulting in a de-duplicated count of 8 net additional pilot cities for the 2016 phase. The final treatment group comprises 49 pilot cities, with non-pilot cities serving as the control group. For pilot city i, if it implemented the pilot policy in year t, the policy–time interaction dummy variable (Did) is specified as 1 for year t and all subsequent years; otherwise, it is set to 0. To rule out confounding from contemporaneous policies, this study employs a text analysis method. All of China’s national, provincial, and urban policy documents in 2011 and 2016 were searched. Keywords covering technology, finance, environmental protection, green, and carbon are used to construct a policy–frequency matrix. No policy mentioning the program under study is found, confirming that the estimated effects are not contaminated by overlapping interventions.

3.2.3. Mechanism Variables

As mentioned earlier, this study employs two mechanism variables. First is the level of technological innovation (Tec). Drawing on the research by Chisht and Sinha (2022) [4], the core purpose of tech-finance policies is to ease financing constraints for technological innovation in general, not to target green innovation specifically. Therefore, focusing solely on green innovation will result in a mismatch between the policy attribute and the variable connotation, leading to estimation bias. Following Sun et al. (2023) [48], we use the number of patent grants per 10,000 people in a region as a measure.

Second is the level of industrial structure upgrading (Uis). Previous studies have generally employed the industrial structure advancement index—specifically, the ratio of added value of the tertiary industry to that of the secondary industry—to represent the level of industrial structure upgrading [49,50]. However, different cities exhibit distinct resource endowments, functional orientations, and specialized sectors, such as agricultural regions tasked with ensuring food security or industrial cities maintaining manufacturing competitiveness. In such regions, growth in tertiary industry output cannot adequately reflect effective upgrading of the regional industrial structure, and focusing solely on the proportion of the tertiary industry has limitations [51]. This study incorporates an additional dimension of industrial structure rationalization to more scientifically and systematically capture the level of regional industrial structure upgrading.

Industrial structure rationalization is characterized by the dynamic advancement of industrial synergy efficiency and inter-industry linkage dimensions. It not only concretely reflects the coordination mechanisms among industries but also directly embodies resource allocation efficiency, functioning as a key benchmark for gauging the coupling degree between factor input structures and output structures [52]. Quantitative studies on industrial structure rationalization remain scarce in the literature. Drawing on the research by Teklie and Doan (2024) [53], we innovatively employ a modified Theil index to measure the degree of industrial structure rationalization across cities. This index offers the advantage of measuring structural deviations between output and employment across different industries, as shown in Equation (2):

$$Thei{l}_{it}={\displaystyle {\sum }_{m=1}^3{y}_{i,t,m}}\ln (y_{i,t,m}/l_{i,t,m}),m=1,2,3$$

(2)

The symbol y_i,t,m represents the added value of the mth industry in the ith region during the tth time period. The variable l_i,t,m represents the proportion of the mth industry’s workforce in the ith region during the tth period relative to the total employment. This Theil index reflects the deviation in the output structure and employment structure of the three major industries in a city. The closer the value is to 0, the smaller the deviation in the industrial structure. Since this index is a negative indicator, it is subsequently subjected to positive data transformation and normalization using the extreme value method. Then, the entropy weight method is used to assign weights to the upgrading of the industrial structure (tertiary industry output/secondary industry output) and the rationalization of the industrial structure (calculated as 0.3185 and 0.6814). Finally, the level of industrial structure upgrading was calculated and constructed based on these two dimensions. The specific algorithms and basic data refer to the study by Zhang et al. (2024) and their database [54].

3.2.4. Control Variables

Informed by existing research [55,56] and ensuring data accessibility, this study incorporates six control variables: urban economic development level (Pgdp), fiscal pressure (Gov), trade openness (Fil), industrialization degree (Ind), transportation infrastructure development (Tra), and education expenditure level (Edu). These variables account for potential confounding factors in analyzing policy impacts.

The detailed operational definitions and measurement methods for the above variables are presented in

Table 2. Qualitative representation of the variables.

Categories	Definitions	Notes	Calculations
Dependent variable	Urban carbon emissions	Ce	Ln (per capita CO2 emissions)
Independent variable	Policy-time interaction term	Did	Dummy variable (1 or 0)
Mechanism variables	Technological innovation	Tec	Number of patent grants per 10,000 people
	Industrial structure upgrading	Uis	Theil index and entropy weight method
Control variables	Urban economic development	Pgdp	Ln (per capita regional GDP)
	Fiscal pressure	Gov	Government general budget expenditure/revenue
	Trade openness	Fil	Total foreign trade volume/regional GDP
	Industrialization degree	Ind	Secondary industry output/regional GDP
	Transportation development	Tra	Ln (regional highway mileage)
	Education	Edu	Education expenditure/general budget expenditure

.

3.3. Samples and Data

To maximize the availability and completeness of the panel data, this study ultimately selected 286 cities in 30 provinces of China, comprising 49 pilot cities and 237 non-pilot cities. The timeframe spans 2005 to 2023, covering 19 years with 5434 observations. The pilot city list was sourced from the official appendices of two policy documents issued by the Ministry of Science and Technology. To measure carbon emissions, we utilized high-resolution data from the Global Atmospheric Emissions Database (EDGAR). Patent data come from the China Research Data Service Platform (CNRDS); supplementary socioeconomic data are from the National Bureau of Statistics database and authoritative publications, including the China Statistical Yearbook. Rigorous data cleaning protocols were implemented. This paper excluded individual samples with important data missing and adopted linear interpolation to impute missing values for the remaining samples [57]. Descriptive statistics for each variable are presented in

Table 3. Descriptive analysis of variables.

Variables	Mean	Standard Deviation	Minimum	Median	Maximum
Ce	10.9714	0.9162	8.1080	11.0217	14.1310
Did	0.1099	0.3134	0.0000	0.0000	1.0000
Tec	0.6491	1.7160	0.0054	0.7339	30.7090
Uis	0.4983	0.1812	0.2561	0.4634	1.8216
Pgdp	10.5135	0.7368	8.1493	10.5617	12.4639
Gov	2.9041	1.9356	0.6490	2.3277	18.4008
Fil	0.2225	0.4082	0.0031	0.8971	2.9230
Ind	0.4618	0.1129	0.1070	0.4652	0.8588
Tra	9.5485	0.7571	6.8459	9.5366	12.1317
Edu	0.1787	0.0437	0.0202	0.1774	0.5916

.

4. Results

4.1. DID Regression Analysis

A DID model (Equation (1)) was used to investigate the causal relationship between tech-finance policies and urban carbon emissions. The model incorporates city-fixed effects and time-fixed effects to mitigate omitted variable bias in panel data analysis. The empirical results are presented in

Table 4. DID regression results.

Variables	(1)	(2)	(3)	(4)
Did	−0.1292 ***	−0.1199 ***	−0.0978 ***	−0.0850 ***
	(−8.8861)	(−8.5427)	(−7.2090)	(−6.5066)
Pgdp		−0.1109 ***	−0.0863 ***	−0.0927 ***
		(−6.3841)	(−4.3008)	(−4.7003)
Gov		0.0202 ***	0.0185 ***	0.0124 ***
		(4.4087)	(4.0816)	(2.6654)
Fil			−1.8388 ***	−1.3695 ***
			(−5.1582)	(−4.2722)
Ind			0.2023 ***	0.1920 ***
			(2.7736)	(2.6330)
Tra				−0.1451 ***
				(−5.2187)
Edu				−0.6400 ***
				(−3.9576)
Constant	10.9860 ***	9.7604 ***	9.9562 ***	11.4032 ***
	(92.4377)	(52.0251)	(50.1880)	(37.3370)
Control variables	Uncontrolled	Part. controlled	Part. controlled	Fully controlled
City-fixed	Y	Y	Y	Y
Time-fixed	Y	Y	Y	Y
R2	0.6760	0.7369	0.7692	0.7801
N	5434	5434	5434	5434

Note: *** denotes significance levels at 1%. Coefficients are reported with t-statistics in parentheses. “Y” indicates control, as below.

, with four progressive specifications.

Column (1) reports baseline estimates without control variables, showing a statistically significant negative coefficient (−0.1292, p < 0.01) for the Did (independent variable). As shown in columns (2) to (4), we sequentially introduce control variables following empirical conventions in environmental economics research. Across all specifications, the core explanatory variable maintains consistent negative coefficients and statistical significance at the 1% level (p < 0.01). These findings confirm that tech-finance policies exert a robust inhibitory effect on urban carbon emissions, and the effect remains significant even after controlling for potential factors. This provides strong empirical support for hypothesis 1.

The relationship between economic development (Pgdp) and urban carbon emissions is negative and has passed the significance test at the 1% level. Thus, the assumption of the latter half of EKC in the theoretical analysis mentioned earlier is reliable. Furthermore, in terms of the significance and direction of other control variable coefficients, fiscal pressure and industrialization degree tend to boost CO₂ emissions, whereas trade openness, transportation development, and education expenditures mitigate urban carbon emissions. These findings align with the literature [58] and offer empirical support for the academic community.

4.2. Parallel Trend Test

Satisfying the parallel trend assumption is an important prerequisite for the DID model. In our quasi-natural experiment, we tested whether there were systematic differences in carbon emissions between the treatment and control groups before the policy was implemented. Referring to the research of Rambachan and Roth (2023) [59], we constructed the following parallel trend analysis model based on the event study method:

$$C{e}_{it}={\alpha }_2+{\displaystyle {\sum }_{t=-6}^6{\beta }_2{D}_{it}}+{\displaystyle \sum \sigma }c{v}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(3)

In Equation (3), D_it is a set of dummies equal to 1 if city i becomes a pilot in year t, and 0 otherwise. All other symbols retain the definitions provided in Equation (1). The estimated coefficients capture the pilot or non-pilot gap in carbon emissions in each relative year and, therefore, merit close scrutiny. We collapse the six pre-policy years into period −6 and the six post-policy years into period +6, with the six pre-policy years as the reference. Figure 2 plots the event–time coefficients. Before the policy, the coefficients fluctuate around zero and cross zero repeatedly, indicating no systematic pre-treatment differences. The samples satisfy the parallel trend assumption and validate the DID design. In multiple periods after policy implementation, the relevant coefficients were significantly negative, further confirming that the implementation of tech-finance policies has a significant inhibitory effect on the carbon emissions of the treatment group.

4.3. Placebo Test

The placebo test is the principal method for endogeneity diagnosis within the DID framework [60]. In this study, we randomly designate pilot cities and assign each a fictitious policy adoption date, thereby constructing a randomized experiment across both the city and temporal dimensions. Using the benchmark specification of Equation (1), we conduct econometric analyses on this artificial sample and determine the distribution of the estimated coefficients. A kernel density approach is employed to implement 1000 random simulations; the resulting distribution is depicted in Figure 3. The coefficients from the placebo regressions are approximately normal, mainly lying between −0.05 and 0.05 and clustering around 0. The true benchmark estimate of −0.085 falls outside the support of the placebo distribution. These findings indicate that our study passes the placebo test, ruling out potential disturbances from other random factors and confirming the absence of endogeneity [61].

4.4. Robustness Tests

We conduct a battery of robustness checks on the DID regression, including propensity score matching (PSM-DID), alternative model, variable substitution, and sample redefinition. The detailed results are reported in

Table 5. Robustness test results.

Variables	(1-a)	(1-b)	(2)	(3)	(4)
	PSM—DID		Alternative Model	Variable Substitution	Sample Redefinition
	Neighbor Matching	Kernel Matching
Did	−0.0363 **	−0.0851 ***	−0.0894 ***	−0.0497 ***	−0.0865 ***
	(−2.348)	(−6.507)	(−5.874)	(−4.438)	(−6.198)
Control variables	Y	Y	Y	Y	Y
Constant	11.7643 ***	11.4028 ***	11.2769 ***	16.1940 ***	11.4439 ***
	(24.455)	(37.333)	(36.469)	(55.196)	(36.822)
City-fixed	Y	Y	Y	Y	Y
Time-fixed	Y	Y	Y	Y	Y
R2	0.6851	0.5558	0.6382	0.7427	0.7770

Note: *** and ** denote significance levels at 1% and 5%, respectively. Coefficients are reported with t-statistics in parentheses. “Y” indicates control, as below.

.

This study employs nearest neighbor and kernel matching to pair each pilot city with non-pilot cities that exhibit the closest observable characteristics. The DID estimates for the matched samples are reported in columns (1-a) and (1-b) of Table 5. The Did coefficients are −0.0363 and −0.0851 under the two algorithms, respectively, both significant at the 5% level. These results confirm that the tech-finance policies’ inhibitory effects on carbon emissions remain robust after accounting for sample selection bias, corroborating the baseline findings.

To test whether the DID model is plagued by the negative weighting issue, this study adopts an alternative model method. We dropped the 2011 policy episode and retained only the 2016 wave. Column (2) shows that after the model is converted from a staggered DID into a single-period one, the coefficient of Did remains significantly negative. Following Wang et al. (2021) and Cook (2024) [62,63], this suggests that negative weighting bias is unlikely to be a salient concern in our setting. In addition, as reported in columns (3) and (4), after adopting variable substitution and sample redefinition, the coefficients of the core independent variable are all negative and significant at the 1% level. The inhibitory effect of Did on carbon emissions remains stable, and the core causal relationship remains robust.

4.5. Heterogeneity Analysis

Research shows that geographic location shapes the extent of regional economic openness and factor mobility [64], urban economic scale influences resource agglomeration and allocation efficiency [65], and resource endowments determine industrial structure and development pathways [66]. Consequently, the same policy is likely to generate markedly heterogeneous effects across regions. Accordingly, this study conducts a heterogeneity analysis along these three dimensions. We aim to provide a scientific basis for subsequent policy formulation and evaluation tailored to local conditions. The results are reported in

Table 6. Heterogeneity analysis results.

Variables	Geographic Location		Economic Scale		Resource Endowments
	Eastern	Central and Western	Large	Small	Resource-Based	Non-Resource-Based
Did	−0.0954 ***	−0.0523	−0.0777 ***	−0.0108	−0.0381 *	−0.1126 ***
	(−5.7360)	(−1.3256)	(−5.5264)	(−0.2772)	(−1.9038)	(−7.1241)
Control variables	Y	Y	Y	Y	Y	Y
Constant	11.6440 ***	11.5389 ***	12.7587 ***	11.2581 ***	12.4237 ***	10.8084 ***
	(22.4626)	(14.8645)	(34.0583)	(23.0666)	(31.3039)	(22.7681)
City-fixed	Y	Y	Y	Y	Y	Y
Time-fixed	Y	Y	Y	Y	Y	Y
R2	0.6492	0.7227	0.7680	0.6749	0.7181	0.6662

Note: *** and * denote significance levels at 1% and 10%, respectively. Coefficients are reported with t-statistics in parentheses. “Y” indicates control, as below.

.

First, we examine heterogeneity across geographic locations. Table 6 shows that tech-finance policies have had a significant effect in eastern Chinese cities, but the carbon reduction effects are statistically insignificant in central and western China. This divergence likely reflects the eastern region’s superior economic foundation and open environment, which foster the intense agglomeration of tech-finance resources and strong innovation vitality among high-tech firms [67]. Within this setting, the policies effectively direct capital to clean-energy R&D, accelerating green technology commercialization and strengthening carbon reduction. For instance, the policies have spurred the rapid expansion of new-energy vehicles and other low-carbon industries that simultaneously bolster economic growth and curb emissions [68]. Conversely, the central and western regions are both major hubs for traditional energy extraction and longstanding recipients of energy-intensive industrial transfers. These areas lag in tech-finance resources and innovative capacity. Although the policies create new development opportunities, constraints arising from limited access to innovation financing, weaker technological capabilities, entrenched industrial structures, and city-specific functional positioning hinder the full carbon reduction potential, resulting in marked geographic heterogeneity in the effectiveness of the policies [69].

Second, we analyze heterogeneity based on economic scale. Urban-economic-scale heterogeneity significantly influences policy implementation effectiveness. Following Bhatti et al. (2023) [70], this study categorizes cities into large-scale and small-scale groups based on the median GDP of the sample, aiming to explore heterogeneity in causal relationships. As shown in Table 6, tech-finance policies significantly curbed carbon emissions in large-scale cities, but the impact is insignificant in small-scale cities, indicating pronounced heterogeneity between the two groups. Large-scale cities, leveraging agglomeration effects, exhibit distinct advantages in allocating tech-finance resources and integrating innovation elements [71]. The concentration of financial institutions, research institutes, and high-end talent provides a solid foundation for implementing such policies, facilitating their conversion into low-carbon technological innovations. By contrast, less developed cities, facing fiscal pressures and prioritizing growth, show limited demand for industrial transformation and green technology adoption, remaining insensitive to the carbon reduction effects induced by tech-finance policies [72].

Third, we explore heterogeneity based on resource endowments. Differences in resource endowments shape cities’ industrial structure and development trajectory, thereby conditioning the carbon reduction efficacy of tech-finance policies [73]. Based on official classification, this study partitions the sample into resource-based and non-resource-based cities. The subsample regressions in Table 6 indicate that tech-finance policies reduce carbon emissions in both groups; however, the effect is stronger and statistically more significant in non-resource-based cities. These cities typically emphasize high-tech manufacturing and modern services, with a relatively low-carbon industrial structure. Consequently, they respond more proactively to tech-finance policies, swiftly channeling policy funds into R&D on energy-saving technologies and the creation of green financial products, which markedly amplifies carbon reduction performance [74]. By contrast, resource-based cities rely heavily on traditional extractive industries with high energy intensity and carbon intensity. Although tech-finance policies provide capital and technological support for industrial transformation, entrenched path dependence and elevated transition costs impede the realization of their full carbon reduction potential [75].

Overall, tech-finance policies exert significantly stronger carbon reduction effects on large, eastern, non-resource-based cities, while the inhibitory effect is markedly weaker in smaller, central, western, and resource-based cities. These findings reveal pronounced heterogeneity in policy effectiveness across cities with differing characteristics.

4.6. Mechanisms Analysis

From the preceding EKC discussion, we conclude that tech-finance policies can reduce urban carbon emissions through their dual roles in promoting technological innovation and industrial structure upgrading. Accordingly, we examined the mechanisms of technological effects and structural effects, with the results presented in

Table 7. Mechanism analysis results.

Variables	Technological Effects		Structural Effects
	Tec	Ce	Uis	Ce
Did	0.3302 ***	−0.0807 ***	0.0709 ***	−0.0741 ***
	(7.017)	(−6.246)	(5.917)	(−5.644)
Tec		−0.0131 ***
		(−2.903)
Uis				−0.1542 ***
				(−8.354)
Control variables	Y	Y	Y	Y
City-fixed	Y	Y	Y	Y
Time-fixed	Y	Y	Y	Y
R2	0.686	0.664	0.632	0.623

Note: *** denotes significance levels at 1%. Coefficients are reported with t-statistics in parentheses. “Y” indicates control, as below.

. Moreover, we conducted a bootstrap test to further examine the significance of these mechanisms, with results in

Table 8. Bootstrap test results for mechanisms analysis.

Mechanisms	c	a	b	a × b	a × b (p-Value)	a × b (95% BootCI)	c′	Effect Ratio (a × b)/c
Did ⇒ Tec ⇒ Ce	−0.085 ***	0.330 ***	−0.013 ***	−0.004	0.031 **	−0.122~−0.039	−0.081 ***	4.706%
Did ⇒ Uis ⇒ Ce	−0.085 ***	0.071 ***	−0.154 ***	−0.011	0.001 ***	−0.144~−0.013	−0.074 ***	12.941%

Note: “c” denotes the coefficient of Did on Ce; “a” represents the coefficient of effect of Did on Tec or Uis; “b” represents the coefficient of effect of Tec or Uis on Ce; “c′” denotes the coefficient of Did on Ce after including Tec or Uis in the model. *** and ** denote 1% and 5% significance levels, respectively.

.

4.6.1. Intermediary Role of Technological Innovation

The original purpose of tech-finance policies was to promote technological innovation. The results in Table 7 confirm this positive effect, indicating that the pilot policies have achieved significant direct goals. By establishing a diversified funding system, these policies inject capital into innovation activities, thereby opening a new pathway for curbing carbon emissions. On the one hand, policy incentives steer venture capital, government subsidies, and green credit toward low-carbon technologies, such as new energy, energy conservation, carbon capture, and carbon storage [76]. On the other hand, risk-sharing mechanisms embedded in the policy lower innovators’ error costs, motivating firms to pursue disruptive low-carbon R&D [77]. To further verify the robustness of the mechanism model, this study employs the variable substitution method, replacing the original Tec (patents per 10,000 people) with a new GTec (green patents per 1000 people). The correlation coefficient between Did and GTec is 0.5198 ***, which remains positively significant at the 1% level.

Table 7 indicates that technological innovation can effectively curb carbon emissions, and this hypothesis has been validated by the academic literature [78,79]. Taken together, the results suggest that technological innovation is an effective mechanism variable. Table 8 also supports this conclusion. The 95% confidence interval obtained from the bootstrap sampling calculation is (−0.122, −0.039), which does not include 0, and the p-value is less than 5%. These results indicate that technological innovation plays a sound intermediary role. Consequently, hypothesis 2 is validated.

4.6.2. Intermediary Role of Industrial Structure Upgrading

Our empirical results demonstrate that tech-finance policies significantly accelerate urban industrial structure upgrading at the 1% level. By reallocating capital, these policies steer the industrial mix toward low-carbon and high-end segments, achieving structural carbon abatement [80]. Policy-directed resources are preferentially channeled into strategic emerging industries that address technological bottlenecks like next-generation information technology. Meanwhile, the policies mandate targeted financing for SMEs, substantially lowering their financing costs and hastening the formation of specialized and sophisticated industrial clusters. Concurrently, the low-interest, low-collateral funds raise the opportunity cost of capital for energy-intensive and highly polluting firms [81]. Under policy constraints and intensified competition, these firms are compelled to shift from energy-intensive to innovation-intensive operations, sustaining the industrial structure upgrade [82].

The positive impact of industrial structure upgrading on carbon emission reduction has been widely recognized by scholars [83,84]. Industrial structure upgrading involves replacing high-emission industries with low-emission ones, high-carbon technologies with low-carbon ones, and fossil fuels with clean energy sources. These effects combine to reduce urban carbon emissions. In summary, combined with the bootstrap test results in Table 8, the intermediary role of industrial structure upgrading is proven, and hypothesis 3 is supported.

Notably, the effect ratios in Table 8 reveal a novel finding. Industrial structure upgrading exerts a significantly stronger effect than technological innovation, both in its direct impact on carbon emissions (|−0.154| > |−0.013|) and in its intermediary effect (12.941% > 4.706%). Scholars have traditionally prioritized developing new green technologies to advance sustainable development [85]. By innovatively comparing the efficacies, this study reveals that industrial structure upgrading plays a far more prominent role than technological innovation. This insight, emerging against a backdrop of research largely fixated on technology, constitutes a potential breakthrough. Therefore, we contend that both academia and practice should pay greater attention to the pivotal role of industrial structure upgrading.

5. Discussion

5.1. Conclusions

Technological innovation is the core driving force of today’s economic development and a key factor in balancing economic efficiency and sustainability. This study adopted a multi-period DID approach to explore the urban carbon reduction effects of tech-finance policies and obtained a series of valuable findings.

To begin with, the study’s primary finding is that the implementation of tech-finance policies has significantly curbed urban carbon emissions. The original purpose of tech-finance policies was to promote financial institutions’ credit support for technological innovation, thereby driving the development of innovation and the economy. However, anchored in the EKC framework [32,33,34,35], this study posits that tech-finance development fosters a low-carbon economy and provides robust empirical validation thereof. After a series of robustness tests, including PSM-DID, alternative model, variable substitution, and sample redefinition, the causal relationship remained intact. This stems primarily from the capacity of technological innovation to reconcile economic growth with environmental objectives [48]. Technological progress contributes over 60% to broad economic growth and nearly 30% to industrial carbon reduction [5]. Accordingly, we demonstrate that tech-finance policies confer pronounced green added value, offering new insights into leveraging policy tools to preserve economic efficiency while achieving environmental objectives.

Subsequently, our further evidence indicates that the carbon-reducing impact of tech-finance policies exhibits pronounced heterogeneity across geographic location, economic scale, and resource endowment. The policies significantly curb urban carbon emissions in the east, whereas the effect is statistically insignificant in the central and western regions. Comparable heterogeneity surfaces when cities are distinguished by economic scale and resource endowment. Overall, the carbon-reducing impact is confined to large, non-resource-based cities in eastern China.

Moreover, a third key finding is that tech-finance policies achieve carbon reductions by enhancing urban technological innovation and accelerating industrial structure upgrading. Grounded in EKC theory concerning technological and structural effects [40,41], our empirical analysis confirms that technological innovation and industrial structure upgrading are the key intermediaries through which tech-finance policies curb carbon emissions.

Finally, and importantly, this study also reveals that industrial structure upgrading plays a far more prominent role than technological innovation. Using a bootstrap test, we quantify the relative contributions of the two mechanism variables and find that industrial structure upgrading exerts a substantially larger effect, both directly on carbon emissions and through its mediating pathway. This is an interesting and possibly counterintuitive insight.

5.2. Practical Implications

Firstly, governments should sustain and refine targeted support for tech-finance initiatives. Our findings demonstrate that the pilot policies directly stimulate technological innovation and generate additional green value, providing empirical justification for stronger commitment. However, significant regional variation demands place-based calibration. Eastern, large, non-resource-based cities merit intensified support to deepen green gains; central–western, small, and resource-based cities require differentiated instruments aligned with local endowments and industrial profiles, avoiding uniform approaches and allowing for flexibility in boosting policy uptake and effectiveness.

Secondly, pilot cities should act as catalytic hubs, spreading the proven benefits of tech-finance across broader regions. Nearest neighbor matching reveals that proximate cities exhibit convergent carbon reduction effects from tech-finance policies. Leveraging this spatial clustering, pilot cities can convert early mover advantages into regional leadership by facilitating industrial relocation and technology diffusion. These initiatives systematically align complementary industrial and resource strengths across neighboring regions. Simultaneously, they channel financial, human capital, and innovation resources toward adjacent areas, thereby expanding tech-finance deployment and carbon mitigation while delivering compounded economic and environmental benefits to a wider set of regions.

Thirdly, preventing policy distortion in tech-finance policies requires a clear understanding of the mechanism through which technological innovation and, especially, industrial structure upgrading drive carbon reduction. At present, tech finance is receiving unprecedented attention and is driven by both strong policy incentives and massive resource inflows. This combination can foster a quantity-over-quality expansion that diverts capital to suboptimal or even counterproductive projects. Such misallocation, together with short-termist speculation, may precipitate systemic financial risks. Governments and financial regulators must reaffirm the public interest orientation of these policies and institute rigorous safeguards against distortion. Continuous guidance should ensure that funds flow to hard technological innovation and genuine industrial structure upgrading, thereby enhancing sustainability outcomes.

Fourthly, financial institutions and other enterprises should seize policy opportunities to enhance their competitiveness and demonstrate CSR through policy dividends [86]. The next five years constitute a critical policy window for tech-finance and a sprint toward carbon peaking for many countries. As key implementers, financial institutions should actively respond to these policy calls by innovating in tech-finance products and service models. They should focus on developing specialized credit products for low-carbon technological innovation and green industrial development, thereby enhancing their legitimacy and competitive advantage. Enterprises, on the other hand, may actively integrate into the sustainability transition by increasing investment in green technology R&D, energy-efficient equipment upgrades, and industrial structure adjustments. While benefiting from preferential financial resources, they should also embed CSR principles into their development strategies to enhance competitiveness in the green low-carbon market [87].

Fifthly, given the high uncertainty of innovation, tech-finance itself carries substantial risks, which in turn shape the sustainability of its green value. From an international comparative perspective, the development of tech-finance in developed countries relies on mature market constraint mechanisms with less risk. In emerging economies such as China, however, tech-finance often expands faster than the supporting institutions, fostering policy-induced capital surges and misallocation. The industrial structure mechanism revealed in this study provides empirical evidence for emerging economies to avoid the quantity-over-quality pitfalls. This not only addresses the gap in the literature regarding emerging economies but also strengthens the study’s practical reference value for these major carbon-emitting countries.

5.3. Contributions

This study advances scholarly research on environmental economics and sustainability transitions through four key contributions, addressing critical gaps in understanding the role of tech-finance policies in balancing economic growth and environmental objectives.

First, the theoretical integration of tech-finance into environmental economics frameworks. While the literature predominantly examines green finance or traditional technology credit [88], this study pioneers the integration of policy-driven tech-finance into the EKC theoretical framework. We advance the literature by demonstrating how policies generate unique green added value. We redefine the role of financial policies in balancing economic growth and ecological integrity and enrich the theoretical framework linking financial instruments to sustainability transitions. In addition, focusing on government-led tech-finance policies rather than general technology credit indicators is another innovative aspect of this study that distinguishes it from existing research.

Second, this study is among the first to systematically document spatial, scale, and resource-based heterogeneity in tech-finance policies’ impacts. Our heterogeneity analysis reveals pronounced variations in tech-finance policies’ carbon reduction efficacy across cities differentiated by geography, economic scale, and resource endowments. Eastern, large, and non-resource-based cities demonstrate significant carbon reductions, underscoring the heterogeneity of policy outcomes shaped by local innovation ecosystems, industrial path dependencies, and institutional capacities. This finding challenges the literature’s assumption of uniform policy effectiveness. For instance, resource-based cities’ reliance on carbon-intensive industries creates structural barriers to green transition, even with policy support. These findings align with calls in environmental economics to internalize regional disparities into policy frameworks, providing granular evidence for place-based environmental governance.

Third, reshaping theoretical understandings of carbon reduction pathways. Contrary to the dominant focus on green technological innovation in sustainability research, our mechanism tests reveal that industrial structure upgrading exerts a significantly stronger mediating effect than previously thought (12.9% vs. 4.7% for technological innovation). Our finding that industrial structure upgrading outperforms technological innovation in carbon mitigation challenges the conventional technology-centric paradigm [89], urging a reorientation toward structural transformation in sustainability research. The study thereby reorients scholarly attention toward structural transformation as a core pathway in environmental economics, particularly within the EKC framework. Additionally, our innovative integrated industrial structure upgrading index (combining the Theil index and entropy weight methods) provides a more nuanced measure than traditional tertiary/secondary industry ratios, capturing both upgrading and rationalization dimensions. This methodological refinement enhances the reliability of sustainability assessments, particularly in heterogeneous urban contexts.

Fourth, methodological rigor in policy evaluation under complex realities. Adopting a multi-period DID design with staggered policy rollouts (2011 and 2016 waves), we address endogeneity concerns inherent in panel data analyses. Various tests, including PSM-DID, parallel trend tests, and placebo simulations, validate the causal inference of tech-finance policies’ carbon reduction effects. These rigorous designs provide causal evidence for policy effectiveness, addressing a critical gap in the policy evaluation literature. More importantly, our bootstrap mechanism analysis quantifies the relative contributions of technological and structural effects, offering a novel analytical toolkit for disentangling complex policy transmission channels.

Collectively, these contributions underscore the transformative potential of policy-driven tech-finance in fostering sustainable development while emphasizing the need for context-sensitive design and structural transformation. By bridging theoretical advancements with empirical rigor, this study enriches environmental economics scholarship and offers actionable insights for policymakers navigating the dual challenges of economic growth and carbon mitigation.

5.4. Limitations and Prospects

The conclusions of this study have certain applicability limits. Their external validity extends to emerging economies similar to China, where a strong center government and disciplined localities ensure policy bite. Developed countries dominated by markets or economically underdeveloped regions may struggle to replicate these findings due to inadequate policy enforcement or inefficient financial systems. Nevertheless, countries like Brazil, India, and Indonesia, which fit the external validity criteria of this study, have consistently ranked high in worldwide carbon emissions. Therefore, our research remains of significant value for global carbon reduction efforts.

Even though we used text analysis to strip out concurrent policies, the long time span and large city sample mean some unobserved policy co-movements may remain. Their presence hinders the identification of the net carbon effect of the tech-finance policy. Meanwhile, slight differences in statistical definitions across the 286 cities could introduce small measurement errors in carbon emissions, reducing estimation precision. Moreover, a robustness check by collapsing the staggered DID into a single wave design was used to crudely ex-post rule out negative weights. Although this practice is still common, the frontier methods include interaction weighting or cohort decomposition that preemptively neutralize the negative weight problem arising from heterogeneous treatment timing. Future work should adopt these updated methods to improve both the precision and the dynamic credibility of the estimates.

This study highlights policy effectiveness disparities between eastern and non-eastern cities but does not deeply explore interactions between tech-finance policies and regional coordination strategies. Future work could use spatial econometric models to analyze knowledge spillover and carbon mitigation synergies across inter-regional industrial chains. Moreover, future research could integrate firm-level microdata to examine how policies influence carbon emissions through corporate innovation decisions and investment preferences. The analysis of welfare and incidence within related research topics should also be more widely discussed.

Despite its limitations, this paper bridges economic principles and sustainability while addressing the emphasis on green finance mechanisms, policy effectiveness, and sectoral transitions. The focus on industrial structure upgrading as a superior driver of sustainability introduces a paradigm shift with implications for both theory and practice. Given that the relevant issue is an active research area, these matters deserve further exploration.