Golden-Edged Dark Clouds: Climate Policy Uncertainty and Corporate Intelligent Transformation

Abstract

Climate policy uncertainty (CPU) poses formidable challenges to global sustainable development and corporate strategic planning, while intelligent transformation is emerging as a pivotal enabler of organizational sustainability. Using panel data from Chinese A-share listed companies between 2011 and 2022, this study investigates the impact of climate policy uncertainty on intelligent transformation. The results indicate that CPU significantly promotes corporate intelligent transformation, a conclusion that remains robust under various sensitivity tests. Government innovation subsidies, enterprise absorption capacity, and enterprise human capital positively moderate this facilitating effect. A heterogeneity analysis reveals that the effect of CPU on intelligent transformation is more pronounced among firms in sci–tech finance pilot zones, regions with high digital financial inclusion, and those led by CEOs with banking experience. This paper contributes to the literature on climate policy uncertainty by examining its role in corporate intelligent transformation, offering actionable strategies for firms to mitigate climate risks while providing policy insights for developing economies to leverage smart technologies in addressing CPU.

1. Introduction

Climate change represents a critical existential threat to humanity [1]. According to China’s National Climate Center, 2023 witnessed over 200 major climate disasters globally, including extreme heatwaves, tropical cyclones, and accelerated sea-level rise [2]. These anomalies have significantly impeded global sustainable development through multidimensional impacts on food security, ecosystem stability, and critical infrastructure [3,4]. In response, China has implemented progressive climate policies like the National Climate Change Adaptation Strategy 2035 and the Interim Regulations on Carbon Emission Trading Administration. However, increasing extreme weather events, geopolitical tensions, and economic pressures have substantially complicated climate governance, thereby elevating climate policy uncertainty [5]. Climate policy uncertainty (CPU) refers to the uncertainties in climate policy instrument selection (e.g., carbon taxes and carbon trading), implementation schedules, and regulatory stringency [6].

Existing studies have demonstrated that climate policy uncertainty generates substantial adverse effects on both macroeconomic and microeconomic dimensions [7]. At the macroeconomic level, CPU hinders low-carbon energy transition [8], consequently suppressing urban green total factor productivity [9]. Moreover, frequent climate policy adjustments reduce investment returns and exacerbate financial market volatility [10]. At the microeconomic level, earnings fluctuations induced by CPU lead to bank credit contraction, thereby intensifying corporate financing constraints and ultimately restraining green innovation investments [11,12]. These findings collectively indicate that CPU essentially creates a multidimensional risk transmission environment. However, little attention has been paid to whether and how firms can rebuild risk resilience through technological means under such circumstances—a critical gap this study aims to address.

With the advent of the Fourth Industrial Revolution (Industry 4.0), artificial intelligence (AI) technology has become deeply embedded in critical sectors such as healthcare, agriculture, retail, and finance [13,14]. This rapid adoption has propelled the global AI industry to a market size exceeding USD 513.2 billion in 2023, reflecting a year-on-year growth rate of 20.7% [15]. Intelligent transformation, powered by artificial intelligence as its core technology, is emerging as a pivotal engine driving macroeconomic development and facilitating corporate strategic upgrading [16]. Specifically, corporate intelligent transformation significantly enhances operational efficiency by reducing information asymmetry [17] and promotes product innovation [18]. Thereby, it achieves carbon emission reduction and ESG performance improvements through strategic optimization [19,20]. This demonstrates that intelligent technologies effectively elevate corporate operational and innovative capabilities by strengthening information acquisition. However, existing research primarily focuses on technological benefits under stable policy environments, with limited understanding of whether intelligent transformation effects can be strengthened amid frequent policy fluctuations.

As the global leader in artificial intelligence patent filings, China has institutionalized systematic support for corporate intelligent transformation through policy innovations. These include establishing national pilot zones for intelligent transformation and innovative development, along with green finance pilot policies [21,22]. Under the dual background of rising climate policy uncertainty and the accelerating global wave of intelligent transformation, CPU may create strategic opportunities for corporate transformation. Empirical evidence demonstrates that CPU enhances the efficiency of capital allocation by curbing excessive investment [23]. Concurrently, CPU drives enterprises to pursue green technological innovation, thereby facilitating the low-carbon transition [24,25].

Using panel data from Chinese firms (2011–2022), this study examines the impact of climate policy uncertainty on corporate intelligent transformation, while investigating the moderating effects of external factors (government innovation subsidies) and internal factors (absorptive capacity, human capital) in this relationship.

The marginal contributions of this study are three-fold. First, while existing research has primarily examined CPU’s effects on stock market volatility, investment efficiency, and green innovation [26,27,28], we extend the analysis of CPU’s microeconomic consequences to the domain of intelligent transformation. Second, existing studies on intelligent transformation have largely focused on drivers such as financial support, scientific output, digital infrastructure, and workforce competency [29,30,31]. This paper enriches the literature on intelligent transformation drivers by conceptualizing CPU as an external shock. Third, we reveal synergistic effects among government innovation subsidies, firm absorptive capacity, human capital structure, and CPU in facilitating intelligent transformation. Additionally, heterogeneous responses to CPU are further documented across varying levels of financial support. These findings provide empirical evidence and policy insights for firms to adapt to CPU-driven challenges strategically.

The paper is organized as follows: Section 2 develops hypotheses regarding the direct impact of CPU on intelligent transformation and its moderating effects. Section 3 details the research design. Section 4 presents the empirical analysis, while Section 5 examines heterogeneous effects. Finally, Section 6 concludes with policy implications and discussion.

2. Research Hypothesis

2.1. CPU and Enterprise Intelligent Transformation

From a risk resilience perspective, CPU drives enterprises to adopt intelligent technologies as a strategic response to enhance adaptive capacity. The adverse effects of CPU on corporate operations manifest through dual channels: input constraints and output volatility. On the input side, frequent climate policy adjustments exacerbate environmental instability, significantly increasing firms’ financing constraints and operational costs [32]. Specifically, heightened policy uncertainty increases the corporate loan default risk [33]. Since financial markets are endowed with anticipatory risk perception, banks respond by raising lending rates and reducing non-performing loans. These risk-hedging behaviors ultimately lead to tightened corporate credit conditions [34]. On the output side, CPU diminishes corporate free cash flow, decreases total factor productivity [35], and markedly increases earnings volatility.

According to the resource-based view (RBV), the intensification of financing constraints and the decline in production efficiency caused by climate policy uncertainty expose enterprises’ resource deficiencies, thereby compelling them to pursue value-creating resources. Intelligent transformation, as a critical pathway for technological upgrading, constitutes a core competitive advantage for enterprises to mitigate CPU. Facing information asymmetry induced by policy volatility, firms are incentivized to increase R&D investments in intelligent technologies. Through the adoption of intelligent technologies, enterprises can enhance policy signal extraction, thereby reducing resource misallocation and lowering production costs [17]. Furthermore, enterprises may prioritize intelligent transformation to strengthen organizational culture, thereby enhancing personnel management and operational efficiency. This strategic focus helps mitigate the adverse effects of CPU on production activities while stabilizing overall productivity.

From an innovation perspective, heightened climate policy uncertainty paradoxically generates greater market opportunities, incentivizing firms to increase investments in intelligent transformation [36]. According to the compulsory emission reduction effect [37] and Porter’s innovation compensation hypothesis [38], properly designed environmental regulations can effectively stimulate corporate innovation. As carbon emissions constitute the primary driver of climate change, climate policies serve as specialized environmental regulations aimed at reducing emissions and promoting sustainable development [39]. Consequently, CPU inherently reflects potential future regulatory shifts in climate-related policies, thereby triggering an innovation compensation effect [40]. Although CPU may temporarily increase production costs, it ultimately drives technological innovation and digital transformation [41]. Enterprises can leverage intelligent technologies to strengthen risk prevention and management by reducing policy-related information asymmetry [42], while optimizing green production processes to enhance product competitiveness [43]. Additionally, these technologies enable real-time pollution monitoring and dynamic emission control optimization, ensuring full compliance with climate policy requirements. Therefore, we propose the following research hypothesis:

H1. 

Climate policy uncertainty promotes the intelligent transformation of enterprises.

2.2. Moderating Effect Analysis

2.2.1. Government Innovation Subsidies

Policy uncertainty generates volatility in business operations, constrains cash flow, and elevates the risk of capital chain disruptions [44]. As CPU intensifies, financial institutions adopt risk-averse strategies that exacerbate corporate financing constraints. Specifically, they reduce credit issuance to firms [45] and assign greater weight to climate risks in credit assessments, thereby amplifying market volatility. To mitigate economic fluctuations, stabilize market operations, and alleviate corporate liquidity constraints, governments typically increase targeted policy subsidies for affected enterprises [46].

Government innovation subsidies accelerate enterprise intelligent transformation amid climate policy volatility through dual mechanisms: the resource compensation effect [47] and the signal transmission effect [48]. Financial support is the most important driving factor for intelligent transformation [22], especially in developing countries. Government subsidies can effectively mitigate and correct the negative externalities of climate policies [49]. Therefore, direct financial support through innovation subsidies helps reduce distortions in corporate investment behavior and alleviates financial constraints. Meanwhile, according to the signal transmission theory, government innovation subsidies possess a distinct “certification effect” [50]. Firms awarded such subsidies are perceived as having strong developmental and investment potential, thereby transmitting positive signals to market investors. This broadens enterprises’ external financing channels, providing financial support for intelligent transformation [51]. Collectively, both the direct fiscal support and the signaling mechanism alleviate R&D funding constraints imposed by CPU, thereby amplifying its stimulative effect on corporate intelligent upgrading. Therefore, the following hypothesis is proposed:

H2a. 

Government innovation subsidies positively moderate the effect of climate policy uncertainty on promoting enterprises’ intelligent transformation.

2.2.2. Enterprise Absorption Capacity

The dynamic capability theory holds that in highly dynamic competitive environments, the competitive advantage derived from an enterprise’s static resources is time-limited and must be sustained through dynamic capabilities—specifically, the continuous reconfiguration of its capabilities and the integration of resources [52,53]. Absorptive capacity, a core dynamic capability, encompasses an organization’s ability to obtain, digest, transform, and apply external knowledge for commercial purposes [54]. Absorptive capacity comprises two key dimensions: potential absorptive capacity (which includes knowledge acquisition and assimilation capabilities) and realized absorptive capacity (which involves knowledge transformation and utilization) [55].

From the perspective of potential capability, absorptive capacity facilitates the optimal allocation of resources (e.g., capital, human capital, and technology) toward intelligent transformation through its resource reconfiguration effect [56]. Specifically, absorptive capacity initially serves an identification function, effectively translating climate policy information into resource allocation signals for intelligent emission reduction. This conversion process strategically elevates the organizational prioritization of smart technologies. Furthermore, enhanced absorptive capacity enables firms to efficiently extract value from external knowledge flows [57], thereby accelerating knowledge capital accumulation during intelligent transformation. As a critical resource foundation for corporate smart transition [58], knowledge capital effectively mitigates the constraining effects of credit restrictions on high-tech investments in uncertain environments. This mechanism ultimately converts the potential coercive effects of climate policy uncertainty into tangible drivers for intelligent upgrading.

From the perspective of realized capability, absorptive capacity enhances organizational learning efficacy [59,60], enabling firms to maintain both advanced smart technology adoption and their innovative capacity despite climate policy uncertainty [61]. As a critical enabler for integrating external intelligent technologies with internal knowledge bases [62], absorptive capacity facilitates knowledge recombination and cross-domain knowledge synthesis. This integrative mechanism expands the scope of corporate smart technology innovation and drives practical implementation. Thus, this paper proposes the following hypothesis:

H2b. 

Absorptive capacity positively moderates the effect of climate policy uncertainty on promoting enterprises’ intelligent transformation.

2.2.3. Human Capital Level

The capital–skill complementarity theory posits that skilled labor exhibits significantly stronger complementary effects with physical capital compared with unskilled labor [63,64]. In other words, highly skilled workers can synergize more effectively with intelligent equipment and technologies, generating technology diffusion effects that enhance marginal productivity [65].

High-level human capital serves as a critical enabler for corporate intelligent transformation [66]. In contexts of frequent climate policy fluctuations, it empowers three key processes: market identification, managerial decision-making, and technology assimilation [67], thereby enhancing organizational resilience to policy risks. First, leveraging specialized technical competencies and policy acumen, skilled workforces can accurately discern structural shifts in the market demand [68]. When climate policies fluctuate, this market-sensing capability further develops into acute environmental perceptiveness toward external complexities. Building on this cognitive advantage, firms strategically allocate more resources to intelligent monitoring and predictive systems, consequently elevating the strategic priority of intelligent transition initiatives. Second, at the management decision-making level, high-quality human capital can enhance intellectual property protection [69], thereby mitigating technology spillover risks and safeguarding technological reserves. Simultaneously, highly educated management teams optimize resource allocation and improve R&D input–output efficiency [70], effectively circumventing policy-induced research decision risks [71]. Consequently, in the context of climate policy changes, advanced human capital enhances both the stability of intelligent technology development and the success rate of corporate transformation. Finally, regarding technology implementation, skilled human capital with strong digital competencies enhances firms’ technology absorption capacity and physical capital utilization efficiency [72]. This capability bridges the gap between strategic planning and the large-scale application of intelligent technologies, ensuring their substantive implementation. Thus, this paper puts forward the following hypothesis:

H2c. 

Human capital upgrading positively moderates the effect of climate policy uncertainty on promoting enterprises’ intelligent transformation.

Based on the aforementioned theoretical analysis, this study establishes a mechanism model (Figure 1), illustrating how climate policy uncertainty influences intelligent transformation. From the resource-based view, external risks drive enterprises to adopt risk-mitigation technologies (particularly intelligent technologies). Consistent with the Porter hypothesis, as a forward-looking indicator of environmental regulation, climate policy uncertainty generates an innovation compensation effect that promotes corporate intelligent transformation. Furthermore, this study identifies three critical moderating pathways. First, building on signaling theory, government innovation subsidies enhance the positive effect of climate policy uncertainty through resource compensation. Second, grounded in dynamic capability theory, organizational absorptive capacity optimizes the transformation process via resource reconfiguration. Third, following capital–skill complementarity theory, human capital levels improve transformation efficiency by facilitating technology diffusion.

To further elucidate the relationship between the core variables, Figure 2 presents the temporal trends of climate policy uncertainty (CPU) and enterprise intelligent transformation (INT) from 2011 to 2022. The figure reveals synchronized upward trajectories for both CPU and INT, with their parallel trends providing preliminary support for Hypothesis H1.

In Figure 2, the solid line (INT) represents the enterprises’ intelligent transformation intensity, measured as the annual sum of AI-related term frequencies in corporate financial reports. The dashed line (CPU) represents climate policy uncertainty, calculated as the yearly summation of prefecture-level climate policy uncertainty indices (see Section 3.2.2 for methodological details). From 2011 to 2022, INT exhibited consistent growth, culminating in a 12-fold increase over the period. This trajectory may stem from either a growing number of firms adopting intelligent transformation or increased implementation intensity among firms. Despite intermittent declines in 2014, 2019, 2020, and 2023, CPU exhibited an overall upward trajectory. The synchronized upward trends of both variables suggest a potential positive correlation between INT and CPU, to be rigorously verified in the subsequent empirical analysis.

3. Research Design

3.1. Empirical Framework

To examine the impact of CPU on intelligent transformation, this study employs a two-way, fixed-effects model as the baseline regression framework. The baseline regression specification is formulated in Equation (1).

$${INT}_{it}={\alpha }_0+{\beta }_1{CPU}_{jt}+{\gamma }_1{Controls}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(1)

${INT}_{it}$ represents the intelligent transformation of enterprise $i$ in year $t$. ${CPU}_{jt}$ is the climate policy uncertainty index faced by prefecture-level city $j$ in year $t$. ${\beta }_1$ captures the net effect of climate policy uncertainty on enterprises’ intelligent transformation. ${Controls}_{it}$ is the set of control variables of enterprise $i$.${\mu }_i$ and ${\lambda }_t$ represent firm and year fixed effects, respectively, and ${\epsilon }_{it}$ is the random error term.

In addition, in order to test the moderating mechanism, this paper constructs the following models (2)–(4).

$${INT}_{it}={\alpha }_0+{\beta }_1{CPU}_{jt}+{\beta }_2{CB}_{it}+{\beta }_3{{CPU}_{jt}\times {CB}_{it}+\gamma }_1{Controls}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(2)

$${INT}_{it}={\alpha }_0+{\beta }_1{CPU}_{jt}+{\beta }_2{EAC}_{it}+{\beta }_3{{CPU}_{jt}\times {EAC}_{it}+\gamma }_1{Controls}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(3)

$${INT}_{it}={\alpha }_0+{\beta }_1{CPU}_{jt}+{\beta }_2{HC}_{it}+{\beta }_3{{CPU}_{jt}\times {HC}_{it}+\gamma }_1{Controls}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$

(4)

Models (2)–(4) add the moderator variable and the interaction term between the independent variable and the moderator based on the benchmark regression model (1). The coefficient ${\beta }_2$ represents the net effect of the moderator on the regression of the dependent variable. The coefficient ${\beta }_3$ measures the moderating effect brought by the moderating variable to the main regression model. ${CB}_{it}$ is the government innovation subsidy obtained by enterprise $i$ in year $t$. ${EAC}_{it}$ is the absorptive capacity of enterprise $i$ in year $t$. ${HC}_{it}$ is the human capital structure of enterprise $i$ in year $t$. The rest are consistent with the benchmark regression model (1).

3.2. Variable Selection

3.2.1. Explained Variables

The intelligent transformation (INT) level: Following the research design of Yao et al. [73], we developed a domain-specific lexicon for intelligent transformation by applying machine learning to the annual report disclosures of listed companies. Through a word frequency analysis and a standardization process on 73 core keywords, we ultimately constructed an enterprise-level index measuring intelligent transformation.

3.2.2. Core Explanatory Variables

Climate policy uncertainty (CPU): We measured CPU using the climate policy uncertainty index at the prefecture level in China. Following the methodological framework of Ma et al. [74], this study employed deep learning algorithms to perform a textual analysis on six mainstream Chinese newspapers, calculating the ratio of climate policy-related articles to the total articles for each outlet. We then averaged these ratios across all the media sources within the same calendar year to construct the final climate policy uncertainty index.

3.2.3. Control Variables

This study selected the following control variables at the firm level to mitigate potential confounding effects:

Ownership type (State): A dummy variable coded 1 for state-owned enterprises (SOEs) and 0 otherwise, accounting for institutional and governance factors that may independently affect intelligent transformation investments.

Firm size (LnSize): The natural logarithm of total assets, controlling for economies of scale to ensure that observed variations in intelligent transformation are not driven solely by size differences.

Firm age (LnAge): The logarithm of years since establishment, controlling for life-cycle effects on transformation capability.

Leverage ratio (Lev): The total debt divided by the total assets. A higher Lev may stimulate intelligent transformation, while a lower Lev may constrain it.

Return on equity (ROEB): ROEB is calculated as the net profit divided by the net assets, which reflects profitability. Firms with a higher ROEB retain more earnings and may be more willing to invest in long-term projects.

Cash ratio (LnCas_r) and current ratio (LnCur_r): These are assessed by the proportion of cash assets and the current assets-to-liabilities ratio, respectively, to evaluate the liquidity risk. Firms with higher liquidity are better positioned to bear the short-term costs of intelligent transformation, which may influence the regression results.

Ownership concentration (LnEC): The largest shareholder’s stake percentage, mitigating the interference of principal-agent conflicts between management and shareholders with the regression results.

Financing constraints (FCs): Measured by the FC index, where higher values indicate a greater difficulty in obtaining external financing. Controlling for FCs helps mitigate the confounding effects between financing capacity differentials and policy sensitivity, thereby facilitating the isolation of the net effect of CPU.

Logarithmic transformations were applied to LnSize, LnAge, LnCas_r, LnCur_r, and LnEC to mitigate skewness and heteroscedasticity.

3.2.4. Moderating Variables

Government innovation subsidy (CB): The logarithmically transformed value of actual government innovation subsidies received by the enterprise during the current fiscal year. This indicator reflects the extent of targeted public sector support for corporate technological upgrading and serves as the key measure for evaluating policy resource transmission efficiency.

Corporate absorptive capacity (EAC): Following the methodology of Yang et al. [75], this study measures absorptive capacity by the R&D intensity, calculated as the ratio of R&D expenditure to the operating revenue, with higher values indicating stronger absorptive capabilities.

Human capital level (HC): The percentage of employees holding bachelor’s degrees or higher qualifications, where increased values correspond to enhanced human capital endowment.

3.3. Data Sources and Descriptive Statistics

The implementation of China’s 12th Five-Year Plan in 2011 substantially intensified carbon reduction policies, leading to persistently rising climate policy volatility. Listed companies’ financial data are publicly transparent and demonstrate significantly better completeness than unlisted firms. Therefore, based on data availability considerations, this study utilizes unbalanced panel data comprising China’s A-share listed companies spanning 2011 to 2022. The firm-level data are sourced from the following authoritative public databases: the CSMAR database, Wind database, and listed companies’ annual reports. Additional data originates from policy reports in mainstream Chinese media, official documents of the Ministry of Ecology and Environment of China, the China Statistical Yearbook, and the China City Statistical Yearbook. To mitigate outlier interference, the sample undergoes the following treatments: (1) the exclusion of financial sector enterprises; (2) the removal of ST and *ST-labeled companies; (3) the application of 1% Winsorization to all continuous variables.

Table 1. Descriptive statistics of variables.

VARIABLES	Obs	Mean	SD	Min	Max
INT	29,437	0.835	1.187	0.000	6.250
CPU	29,940	1.734	0.557	0.503	3.340
State	28,523	0.326	0.469	0.000	1.000
LnSize	28,876	22.019	1.391	19.300	27.100
LnAge	29,940	2.859	0.368	0.000	4.160
Lev	29,313	0.434	0.220	0.051	0.972
ROEB	29,067	0.057	0.164	−0.883	0.397
LnCas_r	28,843	−0.718	1.132	−3.700	2.360
LnCur_r	28,843	0.606	0.772	−1.270	2.860
LnEC	21,819	3.391	0.477	2.100	4.310
FCs	27,800	0.504	0.292	0.002	1.020
CB	21,171	13.948	1.845	0.000	18.000
EAC	26,004	0.042	0.047	0.000	0.285
HC	23,422	0.313	0.222	0.019	0.931

presents the descriptive statistics. The core explanatory variable CPU exhibits a standard deviation of 0.557 and a mean of 1.734, with the standard deviation being substantially smaller than the mean, demonstrating a relatively stable distribution. The dependent variable intelligent transformation (INT) shows a coefficient of variation (the standard deviation-to-mean ratio) of 1.42, exceeding that of the CPU, suggesting significant heterogeneity in digital transformation levels across enterprises. The statistical characteristics of other control variables align generally with findings from prior studies.

4. Empirical Analysis

4.1. Baseline Regression

Table 2. Results of baseline regression.

VARIABLES	(1)	(2)
	INT	INT
cpu	0.030 **	0.038 ***
	(2.55)	(2.76)
State		0.053
		(1.05)
LnSize		0.179 ***
		(8.65)
LnAge		0.022
		(0.13)
Lev		0.061
		(0.61)
ROEB		0.078 *
		(1.87)
LnCas_r		−0.006
		(−0.43)
LnCur_r		0.006
		(0.21)
LnEC		−0.174 ***
		(−3.98)
FCs		−0.115 *
		(−1.86)
_cons	0.778 ***	−2.649 ***
	(38.22)	(−4.01)
Year FE	Yes	Yes
Firm FE	Yes	Yes
N	29,188	19,990
F	6 **	14 ***
r2	0.778	0.772

Notes. Robust t-statistics are reported in parentheses. * p < 0.10; ** p < 0.05; *** p < 0.01.

presents the baseline regression results. All the specifications employ cluster-robust standard errors at the individual level to account for potential serial correlations within the same observational units over time. Column (1) includes only firm and year fixed effects and does not incorporate control variables. The results show that CPU is significant at the 5% level, with a coefficient of 0.030, indicating that a 1% increase in climate policy uncertainty is associated with an approximately 3% increase in the level of enterprise intelligent transformation. Column (2) presents the results after incorporating control variables. The coefficient for CPU remains positive (β₁ = 0.038) and is statistically significant at the 1% level. It shows that climate policy uncertainty significantly enhances the level of enterprise intelligent transformation, supporting Hypothesis H1. On the one hand, the financing constraints and production efficiency pressures intensified by CPU compel enterprises to accelerate technological upgrading. On the other hand, the innovation compensation effect stimulated by it prompts enterprises to actively implement intelligent transformation to build a long-term competitive advantage in policy fluctuations. This finding suggests that governments could moderately raise climate policy uncertainty to promote corporate intelligent transformation.

4.2. Endogeneity Test

To address potential endogeneity issues arising from omitted variables in the model estimation and any measurement bias in the intelligent transformation assessment, this study employed the instrumental variable approach for identification and validation. Areas with dense river networks tend to exhibit higher climate sensitivity and greater volatility in climate policies. Since river density is a time-invariant physical variable, we constructed the instrumental variable Crisk_river by interacting municipal-level river density with the national count of extreme climate events [76]. Frequent extreme climate events enhance governmental urgency for climate governance, thereby increasing the climate policy adjustment frequency. This mechanism ensures the correlation between Crisk_river and climate policy uncertainty. As natural phenomena, both river density and extreme climate patterns remain independent of micro-level corporate behavior, satisfying the exogeneity requirement. Therefore, Crisk_river meets the essential criteria for valid instrumental variables.

The two-stage least squares (2SLS) method was implemented for endogeneity testing, with the regression results presented in

Table 3. Endogeneity test results.

VARIABLES	(1)	(2)
	First Stage	Second Stage
Crisk_river	0.004 ***
	(8.20)
CPU		2.136 ***
		(10.40)
Controls	Yes	Yes
Cragg–Donald Wald F	147.986
Stock–Yogo	16.38
N	15,084	15,084

Notes. Robust t-statistics are reported in parentheses. *** p < 0.01.

. Column (1) displays first-stage regression outcomes, where the instrumental variable (Crisk_river) demonstrates a statistically significant positive coefficient (1% level) on the endogenous variable (CPU), confirming their strong association. The Cragg–Donald Wald F-statistic reaches 147.986, substantially exceeding the Stock–Yogo test’s stringent critical value of 16.38 (10% maximal IV size), thereby rejecting weak instrument concerns. The Durbin–Wu–Hausman test confirms a significant divergence (p < 0.01) between the instrumental variable estimates and the OLS estimates, validating the presence of an endogeneity bias. Column (2) presents the second-stage results, showing a statistically significant positive relationship between CPU and intelligent transformation (INT). These findings indicate that climate policy uncertainty maintains its positive effect on corporate intelligent transformation after controlling for endogeneity, thereby reinforcing the baseline regression conclusions.

4.3. Robustness Test

4.3.1. Modified Estimation Approach

Given the left-truncated nature of enterprise intelligent transformation data, we employed the Tobit model for robustness checks. This methodology effectively addresses any estimation bias caused by data truncation, thereby enhancing the model’s accuracy and reliability. As shown in column (1) of

Table 4. Robustness test results.

VARIABLES	(1)	(2)	(3)
	Tobit	Non-COVID-19	Non-DCM
CPU	0.218 ***	0.073 ***	0.044 ***
	(0.05)	(5.58)	(4.09)
Controls	Yes	Yes	Yes
Cons	−10.596 ***	−2.247 ***	−3.099 ***
	(0.952)	(−4.60)	(−7.50)
Year FE	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes
N	20,164	13,923	19,073
R2		0.762	0.777

Notes. Robust t-statistics are reported in parentheses. *** p < 0.01.

, the coefficient for climate policy uncertainty (CPU) remains statistically positive (β₁ = 0.218, p < 0.01), aligning with the baseline regression direction. These results confirm that climate policy uncertainty retains its significant positive effect on corporate intelligent transformation. This demonstrates the robustness of climate policy uncertainty’s promoting effect on corporate intelligent transformation.

4.3.2. The Exclusion of the Epidemic

The economy experienced significant fluctuations during the COVID-19 pandemic. At the macro level, governments intensified macroeconomic adjustments to address structural imbalances, which consequently amplified policy uncertainties. At the micro level, the pandemic’s short-term shocks exacerbated corporate financing constraints, potentially hindering business development. However, the pandemic-induced decline in the real economy and challenges in supply chain upgrading may also drive enterprises to pursue digital and intelligent transformation [77]. This interplay of multiple factors may confound the identification of significant promoting effects in the benchmark regression analysis, potentially compromising the robustness of the results. Accordingly, we excluded observation samples from 2020 onwards and reran the regression. As shown in column 2 of Table 4, the coefficient of CPU significantly increases from 0.038 to 0.073 (p < 0.01), confirming the robustness of our baseline regression. Moreover, it demonstrates that after excluding the impact of COVID-19, CPU’s enhancing effect on intelligent transformation becomes stronger. In other words, the pandemic shock weakened CPU’s positive influence on intelligent transformation. These results indicate that the pandemic’s economic suppression effects outweigh its generated intelligent transformation opportunities, suggesting enterprises need to establish preventive measures against such global emergencies.

4.3.3. Exclusion of Direct-Controlled Municipalities

Given that municipalities directly under the central government exhibit systematic differences from ordinary prefecture-level cities in policy implementation, resource allocation, and labor structure, enterprise intelligent transformation levels may be influenced by municipal-level factors. This study excluded samples of enterprises registered in the four central municipalities (Beijing, Shanghai, Tianjin, and Chongqing) to conduct robustness tests. The estimation results are presented in column (3) of Table 4. The coefficient of CPU is significantly positive (β₃ = 0.044, p < 0.01), consistent with the baseline regression results. This indicates that climate policy uncertainty still significantly promotes corporate intelligent transformation in non-municipality regions. These findings demonstrate that the significant promotion effect of policy volatility on intelligent transformation is not influenced by regional administrative characteristics, confirming the robustness of the results.

4.4. Moderation Effect Test

4.4.1. Government Innovation Subsidies

The regression results presented in

Table 5. Adjustment results of government innovation subsidies.

VARIABLES	(1)	(2)
	INT	INT
CPU	0.038 ***	−0.109
	(2.76)	(−1.29)
CB		−0.014
		(−1.19)
CPU×CB		0.011 *
		(1.83)
Controls	Yes	Yes
Year FE	Yes	Yes
Firm FE	Yes	Yes
N	19,990	14,965
R2	0.772	0.776

Notes. Robust t-statistics are reported in parentheses. * p < 0.10; *** p < 0.01.

show that the coefficient estimate for the interaction term CPU × CB (β₃ = 0.011) in column (2) is statistically significant at the 10% level. It indicates that for each additional unit of government innovation subsidy received by enterprises, the positive impact of climate policy uncertainty on intelligent transformation increases by 0.011 units. These results demonstrate that government innovation subsidies play a significant positive moderating role in CPU’s promotion effect on enterprise intelligent transformation, thereby validating Hypothesis H2a. This moderating effect primarily stems from the “resource compensation effect” and “signal transmission effect” of government innovation subsidies. The results carry an implicit policy implication: governments should couple increased climate policy uncertainty with augmented innovation subsidies for enterprises.

4.4.2. Enterprise Absorptive Capacity

The moderating effect results are presented in

Table 6. Moderating results of corporate absorptive capacity.

VARIABLES	(1)	(2)
	INT	INT
CPU	0.038 ***	0.015
	(2.76)	(0.92)
EAC		−0.575
		(−0.82)
CPU×EAC		0.596 **
		(2.02)
Controls	Yes	Yes
Year FE	Yes	Yes
Firm FE	Yes	Yes
N	19,990	18,282
R2	0.772	0.871

Notes. Robust t-statistics are reported in parentheses. ** p < 0.05; *** p < 0.01.

. As shown in column (2), the coefficient β₃ of CPU×EAC is 0.596 and statistically significant at the 5% level. It demonstrates that absorptive capacity exerts a significant positive moderating effect on CPU’s promotion of intelligent transformation. In other words, enterprises with a lower absorptive capacity exhibit weaker promotion effects from CPU, while those with a higher absorptive capacity can more rapidly and effectively enhance their intelligent transformation to respond to climate policy uncertainty. Hypothesis H2b is thereby validated. By acquiring external beneficial information and enhancing organizational learning resilience, absorptive capacity transforms market opportunity signals from climate policy uncertainty into drivers of intelligent transformation. Consequently, enterprises should establish dynamic learning mechanisms to enhance absorptive capacity, thereby synergistically facilitating intelligent transformation.

4.4.3. Human Capital Level

The moderating effects are presented in

Table 7. Adjustment results of human capital upgrading.

VARIABLES	(1)	(2)
	INT	INT
CPU	0.038 ***	−0.014
	(2.76)	(−0.52)
HC		−0.076
		(−0.41)
CPU×HC		0.201 ***
		(2.74)
Controls	Yes	Yes
Year FE	Yes	Yes
Firm FE	Yes	Yes
N	19,990	15,512
R2	0.772	0.783

Notes. Robust t-statistics are reported in parentheses. *** p < 0.01.

. As shown in column (2), the coefficient of the interaction term CPU×HC is significantly positive (β₃ = 0.201, p < 0.01). This indicates that the human capital level positively moderates CPU’s positive impact on enterprise intelligent transformation, confirming Hypothesis H2c. Specifically, for every one-unit increase in a firm’s human capital level, the positive effect of climate policy uncertainty on intelligent transformation strengthens by 0.201 units. A high-quality workforce enhances firms’ risk–technology response efficiency through improving market identification capabilities, optimizing management decisions, and facilitating technology implementation. It extends empirical evidence for the capital–skill complementarity theory in the context of CPU and intelligent transformation.

5. Heterogeneous Impact of Financial Support

Adequate financial support can mitigate the adverse effects of climate policy uncertainty by enhancing risk hedging capabilities and technological innovation investments, thereby resulting in the differential sensitivity of enterprise intelligent transformation to climate policy changes across varying financial conditions [78]. This study examines the heterogeneous impacts of financial support from three dimensions: technology finance, digital finance, and CEO banking experience.

5.1. Sci–Tech Finance

The pilot policy for science and technology finance aims to foster new quality productivity by promoting the adaptive restructuring of technological elements and financial resources through policy experimentation. This study constructed the sci–tech finance pilot policy variable using the interaction term between the “policy implementation timing” (time) and “pilot cities” (treat). The sample was divided into enterprises located in science and technology finance pilot cities and those in non-pilot cities for the subgroup analysis. Columns (1)–(2) in

Table 8. Heterogeneity test results.

VARIABLES	(1)	(2)	(3)	(4)	(5)	(6)
	Sci–Tech Finance		Digital Finance		CEO Bank Exp.
	Yes	No	Yes	No	Yes	No
CPU	0.068 **	0.007	0.046 ***	−0.001	0.053 ***	0.079
	(2.44)	(0.47)	(3.08)	(−0.04)	(3.54)	(0.97)
Controls	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes
N	7094	12,872	18,186	1699	16,818	346
R2	0.788	0.772	0.776	0.711	0.777	0.887

Notes. Robust t-statistics are reported in parentheses. ** p < 0.05; *** p < 0.01.

demonstrate that the CPU coefficient for pilot cities is 0.068 (p < 0.05). This coefficient shows that in the pilot cities of science and technology finance, for every one unit increase in climate policy uncertainty, the emphasis on intelligent transformation of enterprises increases by 0.068 units. However, the coefficient of non-pilot cities is not significant. These results suggest that the sci–tech finance pilot policy alleviates the capital constraints induced by CPU through improved financing channels and reduced technological R&D risks. Specifically, pilot cities employ specialized financial instruments to fund enterprise intelligence projects, mitigating capital mismatch risks and enhancing corporate financial resilience, thereby significantly amplifying CPU’s incentivizing effect on intelligent transformation investments. Furthermore, industry–finance linkage platforms in pilot areas reduce bank–enterprise information asymmetry via government credit guarantees, directing capital flows toward strategic emerging fields, such as intelligent technologies. In contrast, non-pilot areas lack targeted policy support, making it challenging for enterprises to convert external policy pressures into sustained technological investment commitments. This heterogeneous result shows that sci–tech finance policy can hedge the negative impact of climate policy uncertainty and provides a breakthrough point for resolving the contradiction between environmental governance and technology upgrading through financial instruments under the “dual carbon” goal.

5.2. Digital Finance

We assessed digital finance using the Urban Digital Inclusive Finance Index, published by the Institute of Digital Finance at Peking University. Based on the median value of city-level digital finance development, the sample is categorized into two groups: enterprises located in cities with high versus low levels of digital inclusive finance. The regression results are presented in columns (3)–(4) of Table 8. For enterprises in highly digitally inclusive finance cities, the CPU coefficient is 0.046 (p < 0.01). This coefficient shows that in cities with high-level digital inclusive finance development, for every one unit increase in climate policy uncertainty, the practical application depth of enterprise intelligent transformation technology increases by 0.046 units on a marginal basis. However, the coefficient of the low-level group is not significant. The results indicate that the degree of digital inclusive finance exerts heterogeneous effects on the intelligent-transformation-promoting effect of climate policy uncertainty. From an external support perspective, cities with advanced digital finance reduce the frictional costs of intelligent transformation by consolidating digital financial infrastructure. From an internal absorption capacity perspective, digital penetration enables firms to develop stronger independent R&D capabilities. Conversely, the low-level group struggles to meet rapidly changing financing demands due to rigid approval processes and regional fragmentation in traditional financial services. Therefore, while elevating climate policy uncertainty, governments should vigorously develop digital finance to synergistically promote corporate intelligent transformation.

5.3. CEO Banking Experience

Based on whether the CEO possesses banking experience, we partitioned the sample into two subsamples. As presented in columns (5)–(6) of Table 8, the CPU coefficient shows a statistically significant positive value (β₁ = 0.053, p < 0.01) for firms with banking-experienced CEOs. This coefficient means that when the climate policy uncertainty index increases by one unit, the implementation intensity of the enterprise intelligent transformation strategies of banking-background CEOs increases by 0.053 units. However, the regression coefficient for firms with CEOs lacking banking experience is insignificant. These findings suggest that executives’ financial background exerts heterogeneous effects on regression outcomes. Climate policy uncertainty triggers financial institutions to reduce the credit supply, exacerbating corporate financing constraints. CEOs with banking experience may leverage their industrial networks and credit capital to expand corporate financing channels. Conversely, firms led by CEOs without banking experience might fail to seize policy-driven technological upgrading opportunities due to risk-aversion tendencies or insufficient resource integration capabilities. This finding indicates that in the context of accelerated climate policy adjustment, companies can build policy responsiveness by introducing senior executives from the financial sector.

6. Conclusions and Recommendations

6.1. Conclusions

As climate issues become increasingly severe, the importance of climate policies has correspondingly grown. Against the backdrop of accelerated restructuring in global climate governance systems and the concurrent industrial intelligence transformation, climate policy uncertainty (CPU) has become an unavoidable factor in corporate strategic decision-making. Using a sample of A-share, non-financial listed companies from the Shanghai and Shenzhen stock exchanges between 2011 and 2022, this study examines the effects, mechanisms, and heterogeneous impacts of financial support on how CPU influences corporate intelligence transformation levels.

The main research findings are as follows:

(1)

CPU significantly enhances the level of enterprise intelligent transformation, and the results remain robust after addressing endogeneity concerns, replacing regression models, excluding pandemic effects, and removing municipalities directly under the central government. This suggests that enterprises significantly increase their investment in intelligent transformation technologies to mitigate the risks associated with climate policy adjustments.

(2)

The increases in government innovation subsidies, improvements in enterprise absorptive capacity, and the accumulation of human capital collectively positively moderate CPU’s promotional effect on intelligent transformation. The influence mechanisms manifest through three pathways: First, government innovation subsidies alleviate corporate financial constraints via dual mechanisms of resource compensation and signal transmission. Second, organizational absorptive capacity facilitates the effective integration of internal and external information through dynamic capability theory. Third, human capital levels enhance technological absorptive capacity through capital–skill complementarity effects. The synergistic interaction between CPU and these three factors amplifies the promotion effect on intelligent transformation.

(3)

The heterogeneity analysis revealed that enterprises in technology–finance pilot cities, those with higher digital finance development levels, and those led by CEOs with banking experience demonstrate stronger responsiveness of intelligent transformation levels to CPU.

6.2. Policy Implications

Under the climate policy uncertainty, we propose policy recommendations for collaborative efforts between enterprises and governments to advance intelligent transformation.

First, enterprises should fully utilize intelligent transformation technologies to enhance climate risk response capabilities. At the decision-making level, enterprises should integrate intelligent algorithms into policy monitoring systems to construct a climate risk early-warning matrix containing core parameters, such as carbon price fluctuations and regulatory intensity, thereby shortening the policy response cycle. In terms of production processes, enterprises should apply digital twin technology to establish supply chain collaboration platforms, achieving dynamic optimization of resource allocation to effectively buffer production volatility. Furthermore, corporate intelligent transformation requires systematic policy support from the government. Therefore, the government should establish a special risk compensation fund for intelligent transformation, using financial instruments such as subsidized loans to hedge against market risks in corporate technology upgrades. Simultaneously, it should expand the scale of AI innovation pilot zones to promote the regional adoption and application of intelligent response technologies.

Second, the government should improve the precision and transmission efficiency of smart technology innovation subsidies. On the one hand, the hierarchical evaluation mechanism of intelligent technology maturity should be established to focus on providing stepwise subsidy support to enterprises in the critical stage of industrialization. On the other hand, the government subsidy qualification certification should be embedded in the ESG information disclosure framework of enterprises to correct the short-term investment behavior caused by climate policy fluctuations. At the same time, financial institutions should be encouraged to establish a “white list” rapid approval mechanism based on the rating of the intelligent transformation of enterprises to effectively alleviate the financing constraints of R&D investment.

Third, enterprises need to enhance the resilience of intelligent transformation through organizational innovation and talent system optimization. For the upgrading of organizational structure, enterprises should set up full-time policy analysis teams to track industrial policy trends in real time, dynamically adjust R&D budgets and technology upgrading paths, and ensure the accurate allocation of resources. For the construction of talent ecology, the government should take the lead in formulating an intelligent talent ability framework and clarify the core skill requirements of technology–industry integration. At the same time, the government should improve the service mechanism for foreign experts, give tax incentives to enterprises that introduce top talent in the field of intelligence, and promote localized technological innovation.

Fourth, the government should build a diversified financial support system to ensure the sustainability of enterprise intelligent transformation. The government is advised to establish a science and technology credit risk compensation mechanism and a special fund for intelligent transformation and to support the transformation and application of key technologies through blockchain storage technology. Financial institutions are mandated to rely on supply chain finance platforms to provide differentiated financing solutions to focus on alleviating the liquidity pressure of micro-, small-, and medium-sized enterprises. Corporate entities are required to incorporate the financial strategy into the decision-making system, carry out regular training on capital market operations for the management, and cooperate with professional institutions to formulate financing plans for the window period of technology upgrading.

6.3. Limitations and Future Research Directions

While this study advances the understanding of how climate policy uncertainty promotes corporate intelligent transformation, certain research limitations persist due to inherent methodological and conceptual constraints.

First, due to limitations in data timeliness and model constraints, this study primarily focuses on the short-term responses of enterprises under climate policy uncertainty. However, given the lagged coupling between policy iterations and technological updates, future research should expand the exploration of CPU’s long-term dynamic technological effects. Specifically, further studies could introduce threshold regression models to identify critical points in CPU’s impact on enterprise intelligent transformation and examine the differential effects of uncertainty intensity and policy implementation pace on technological change.

Second, this study primarily employs text analysis methods to measure CPU, which can effectively capture regional variations in CPU. However, more precise variable measurement approaches remain to be explored. Future research could differentiate between “ambiguity” and “stringency” in climate policy texts, quantifying the semantic complexity of policies to enhance the reliability and robustness of empirical tests.

Finally, this study primarily examines the moderating effects of capital, knowledge, and talent, as well as the heterogeneity of financial support, while leaving the underlying mechanisms unexplored. Future research could further explore the transmission mechanisms through which enterprises convert policy pressure into technological innovation momentum. Additionally, enriching the heterogeneous characteristics of CPU’s influence on corporate intelligent transformation represents a promising direction for future research innovation. For instance, comparative studies between high-carbon industries and low-carbon service sectors, or between state-owned enterprises’ “policy priority” and private firms’ “market adaptation” strategies, may reveal potential biases in research findings.