1. Introduction
Carbon emissions have become a critical issue at the national level over the years, due to their contribution to local air pollution, acid rain, and climate change. In China, which is the world’s largest carbon emitter, the National Development and Reform Commission of China launched three rounds of Low-Carbon City Pilot (LCCP) policies in 2010, 2012, and 2017, as a critical step toward its national carbon goals. The cities involved in the LCCP are critical in setting appropriate emission targets, assessing emission indicators, and testing new policies to promote low-carbon industries and technologies to meet local needs. Consequently, local governments of these cities face intensified decarbonization pressures. However, balancing decarbonization with high-quality development remains a significant challenge. Recent studies highlight industrial intelligence—integrating big data analytics, AI, and automation across production processes to enable data-driven, autonomous decision-making and adaptive self-learning—as a promising solution for improving energy and environmental performance [
1,
2]. Given the pivotal significance of industrial intelligence in global technological competition [
3,
4,
5], it is essential to explore whether local governments, under decarbonization pressure, will utilize it to simultaneously achieve decarbonization and high-quality development. Answers to this question provide policymakers with valuable insights for formulating effective low-carbon transformation strategies, thereby advancing high-quality development and addressing global climate imperatives.
Different from previous studies that focus on the effects of LCCP policies, this study employs the policy as an instrument of local governments’ decarbonization pressures, and examines how decarbonization pressures influence their behavior. Using panel data from 283 Chinese cities between 2007 and 2019, the LCCP policies are leveraged as a quasi-natural experiment, and a staggered DID specification and a Causal Forest model are employed to examine the impact of local government decarbonization pressures on urban industrial intelligence. The analysis reveals an increase of 1.96 industrial robots per 10,000 employees after the implementation of the policies, indicating that local government decarbonization pressures significantly boost urban industrial intelligence. Mechanism analysis shows that local government decarbonization pressures promote urban industrial intelligence by encouraging intelligent policies and enterprises’ source-based environmental governance.
Potential endogeneity concerns are addressed through several approaches. The parallel trends assumption is verified with a dynamic DID regression. Selection bias and unobserved heterogeneity are mitigated through the PSM-DID method. A series of robustness tests are conducted to ensure the reliability of the results. Placebo tests confirm that the findings are not driven by any time trends. Regressions are rerun with Callaway and Sant Anna’s [
6] estimators, and additional checks are conducted by replacing the key explanatory variable to avoid bias caused by potential measurement error, and by controlling for confounding policies during the same period.
Additionally, China’s 1994 tax-sharing reform, which transferred major taxes like VAT to the central government, significantly cut local revenues [
7]. However, local governments still must fulfill various high-quality development objectives, creating a mismatch between constrained revenues and escalating expenditures, highlighting the need to explore the moderating role of fiscal pressure on how local government decarbonization pressures impact urban industrial intelligence. The results indicate that fiscal pressure weakens the effect of local government de-carbonization pressures on urban industrial intelligence. Specifically, each additional unit of fiscal pressure reduces the effect by 1.89 industrial robots per 10,000 employees.
Econometric models excel in interpretability and robustness but often fail to capture heterogeneous treatment effects. Causal Forests, combining Random Forests with causal inference, address this limitation by uncovering nonlinear relationships and identifying individual effects across subgroups [
8]. Thus, Causal Forests are further employed to identify variable importance and analyze heterogeneity. An inverted U-shaped relationship is revealed between openness to foreign trade and the Conditional Average Treatment Effect (CATE), while a positive correlation between labor costs and CATE is also identified. Lastly, the decarbonization impact of industrial intelligence is assessed and confirmed to be positive.
The research mainly contributes to the literature on the relationship between local government decarbonization pressures and urban industrial intelligence. Research on industrial intelligence has primarily examined its impact on employment [
9,
10,
11,
12,
13,
14,
15], and a few studies have explored the decarbonization effects of industrial robots [
1,
16]. However, the extent and mechanisms of local governments’ role in the development of industrial intelligence, particularly under decarbonization pressures from higher-level authorities, remain understudied in the literature. The work of Zhang and Zhang [
17] is the most closely related to this paper. They examined how environmental regulation induces firm digitalization. Different from them, the study focuses on the behavior of the local government, examining how local government characteristics influence policy effectiveness.
The second contribution is that enterprise environmental governance is differentiated into source-based and end-of-pipe governances, an area rarely touched on in existing studies. Sun et al. [
18], closely related to this work, classified corporate environmental investments into source-based and end-of-pipe types, showing that green mergers and acquisitions only promote source-based investment. However, different from their study, this study explores how these two governance mechanisms mediate the impact of local government decarbonization pressures on urban industrial intelligence.
Lastly, this paper also contributes to the literature by applying the Causal Forest model, an advanced machine learning technique, to examine the heterogeneous effects of local government decarbonization pressures on urban industrial intelligence.
The remainder of this paper is organized as follows.
Section 2 presents the hypotheses development,
Section 3 describes the model setting and variable selection,
Section 4 discusses the empirical results,
Section 5 examines heterogeneity using Causal Forests,
Section 6 tests the decarbonization effects of industrial intelligence, and
Section 7 concludes with policy implications.
2. Hypotheses Development
2.1. China’s Pressure-Based System and Local Government Decarbonization Pressures
China’s unique “pressure-based system” is characterized by a quantitative task segmentation method with hierarchical breakdowns, a materialized multi-level evaluation system, and a political and economic reward–penalty mechanism [
19]. This system uses evaluation outcomes as criteria for officials’ appointments and promotions, prompting various government levels to fulfill assigned tasks and targets under stringent evaluation-driven pressure [
20].
While local governments face pressures from multiple sources, such as the top-down performance evaluation pressure exerted by higher-level governments on local governments, the horizontal competition pressure among neighboring cities, and the bottom-up demand pressure from citizens on local government, the core mechanism behind these pressures remains “politicization”—higher-level governments, particularly the central government and various Party committees, designate certain key development tasks as “political tasks” and motivate lower-level governments and functional departments to achieve these tasks through political promotions or economic incentives [
21].
Following the economic reforms, economic development was established as a central political directive and became the foremost “political task”. Under the policy of prioritizing economic development, China has used GDP growth rates as a key indicator of political performance for decades, prompting local governments to engage in a “GDP-centered promotion tournament” driven primarily by promotion incentives. This tournament model provided local officials with strong incentives to promote local economic growth, contributing to what has been described as China’s “economic growth miracle”. However, it also resulted in significant resource wastage and environmental degradation, becoming one of the institutional foundations of China’s extensive growth model [
19].
To shift away from an extensive economic growth model and reduce high carbon emissions, China has gradually introduced greenhouse gas emission reduction targets and the “dual carbon” strategy. As a result, carbon reduction has become an increasingly important “political task”. This change has brought about a rising number and greater weight of indicators related to energy conservation, emission reduction, and carbon reduction within the performance evaluation system, as the system evolves from a single-metric focus to a multi-goal framework, including sustainable economic development, social welfare improvement, social progress, and ecological civilization. It has ultimately resulted in growing decarbonization pressures for local governments.
2.2. Local Government Decarbonization Pressures and Urban Industrial Intelligence
Local government officials, driven by the prospect of promotion through achieving environmental governance goals, are incentivized to transfer their decarbonization pressures to other economic entities within their jurisdictions. Consequently, as key players in market economic activities and major sources of carbon emissions, enterprises become the primary recipients of these pressures. Furthermore, enterprises are central to driving industrial intelligence, with the transformation and upgrading of their production systems being crucial for industrial advancement and gaining technological leadership. Under the influence of local government decarbonization pressures, the enthusiasm and initiative of enterprises in pursuing intelligent development directly affect the progress and outcomes of overall urban industrial intelligence.
Figure 1 illustrates the mechanisms through which local government decarbonization pressures affect urban industrial intelligence from the perspectives of both “proactive” and “passive” industrial intelligence of enterprises. Proactive industrial intelligence denotes enterprises’ voluntary adoption of intelligent practices in response to supportive policies, while passive industrial intelligence reflects compelled adoption under regulatory pressure to achieve emission reduction and sustainability goals.
First, under decarbonization pressures, local governments proactively promote various low-carbon technologies, including energy-saving and efficiency-enhancing technologies, non-fossil energy technologies, alternatives for fuels and raw materials, carbon capture, utilization, and storage technologies, and carbon sink technologies. Among these, intelligent technology is a crucial subset of energy-saving and emission-reduction technology. Recognizing that enhancing industrial core competitiveness through industrial intelligence has become a global consensus and trend, local governments prioritize intelligent technology as a primary low-carbon solution, thereby introducing more intelligent supportive policies, including financial incentives, policy consulting, expedited approvals, expanded financing, the designation of model intelligent enterprises, and industry–academia cooperation. These measures encourage enterprises to proactively adopt intelligent solutions to achieve decarbonization, ultimately contributing to urban industrial intelligence at the macro level.
Second, as suggested by the Porter Hypothesis [
22], local government decarbonization pressures strengthen environmental regulations, raising compliance costs for enterprises and “forcing” them to decarbonize through internal environmental governance passively [
22]. Some enterprises may pursue source-based governance, which primarily relies on intelligent products and technologies, to optimize production processes, while others may resort to lower-cost end-of-pipe governance by installing pollution control devices like desulfurization and dust removal systems [
23]. However, in response to local governments’ long-term focus on ecological governance performance, an increasing number of enterprises are moving toward source-based governance to achieve decarbonization and avoid penalties, thereby advancing urban industrial intelligence.
Based on the above analysis, the following hypothesis is proposed:
H1: Local government decarbonization pressures significantly promote urban industrial intelligence.
2.3. The Moderating Effect of Local Government Fiscal Pressure
Amid slowing economic growth, unsustainable land-based financing, and rising local debt, local governments in China are under increasing fiscal pressure.
Due to the constraints of short official tenures and the long investment recovery periods associated with industrial intelligence, local governments may exhibit shortsighted behavior in governance. This can manifest in a “race to the bottom” strategy, by reducing investments in environmental management and regulation to cut fiscal expenditures [
24]. This approach indirectly lowers carbon reduction costs for highly polluting enterprises by diminishing their motivation to cut emissions. Consequently, the intended “passive” industrial intelligence transformation mechanism for enterprises becomes ineffective. This may even “crowd out” corporate investments in intelligent technologies [
25], thereby hindering the advancement of industrial intelligence in urban areas. Hence, the following hypothesis is further proposed:
H2: Fiscal pressure diminishes the positive effect of local government decarbonization pressures on urban industrial intelligence.
7. Conclusions
This paper studies how local government decarbonization pressures have affected urban industrial intelligence. Employing a staggered DID approach, it is found that local government decarbonization pressures significantly promote urban industrial intelligence by encouraging local governments to implement intelligent policies while compelling enterprises to adopt source-based environmental management practices, and fiscal pressure moderates the effect negatively. A Causal Forest model reveals an inverted U-shaped relationship between openness to foreign trade and CATE, while labor costs show a positive correlation with CATE. Additionally, it is also confirmed that urban industrial intelligence effectively contributes to decarbonization.
Our results have the following policy implications. First of all, the central government’s performance appraisal system should be enhanced by incorporating specific indicators—such as carbon emission intensity, the share of renewable energy, and the proportion of green buildings—and by linking evaluation outcomes to low-carbon development performance. Incentives such as targeted fiscal support, tax relief, and preferential financing should be provided to encourage local governments to adopt proactive emission reduction measures. For example, Hangzhou has introduced carbon-linked fiscal mechanisms to foster green innovation. Furthermore, local governments should align intelligent technology initiatives with national strategies—such as the dual carbon targets and industrial digitalization—through investment in industrial intelligence and urban innovation. Effective coordination between central and local authorities is essential to prevent fragmented governance and ensure coherent progress. Lastly, enterprises should develop and adopt intelligent source-based emission reduction technologies to optimize production processes, lower emissions, and accelerate intelligent transformation. For instance, real-time carbon monitoring systems, piloted in Jiangsu’s industrial parks, have effectively identified emission hotspots and guided targeted mitigation efforts. Nonetheless, insufficient coordination, limited technological capacity, or weak incentives may undermine these efforts, increasing the risk of policy failure and regional disparities. Thus, future policies should incorporate adaptive mechanisms to manage uncertainty and support a resilient low-carbon transition.
Despite its contributions, this study has the following limitation: the environmental governance data are limited to listed enterprises, which may not fully represent small and medium-sized enterprises (SMEs). Due to significant differences in governance structures, resource allocation, and environmental practices between listed enterprises and SMEs, the findings may not be broadly applicable. Furthermore, while the study emphasizes the benefits of industrial intelligence, it is important to also acknowledge potential downsides, such as technological displacement, employment precarity, and rising inequality. These challenges must be considered when designing policies that promote industrial intelligence. Future research could address the limitation by incorporating SME data to improve the representativeness and generalizability of the results, as well as exploring the trade-offs between industrial intelligence and its socio-economic impacts.
Although this study centers on China, its findings offer broader implications for local-level decarbonization. Unlike China’s centralized governance model, EU municipalities and U.S. local authorities operate within more decentralized, participatory frameworks. EU cities, for instance, participate in transnational initiatives like the Covenant of Mayors, promoting bottom-up commitments and peer benchmarking. In the U.S., cities such as New York and San Francisco have advanced ambitious climate agendas independently of federal mandates. In contrast, Chinese local governments respond primarily to central directives and performance-based incentives. The alignment of top-down pressure with technological upgrading—particularly through industrial intelligence—may inform innovation-driven climate governance in other institutional settings.