The Evolution of Corporate Shadow Banking Behavior Under Climate Risk: Insights from Resilience and Capital Structure

Abstract

In the context of green transformation, climate change and its economic implications are attracting increasing attention. Based on the Trade-off Theory framework, this study examines how climate risk affects firms’ shadow banking activities in emerging markets. This study focuses on emerging market economies, using a panel dataset of Chinese A-share non-financial listed firms from 2007 to 2023 to systematically examine the relationship between climate risk and shadow banking activities, that is, financing conducted outside the formal banking system. The empirical findings reveal that climate risk significantly dampens the shadow banking activities of non-financial firms. Further mechanism analysis suggests that this effect operates through two key channels: the weakening of corporate resilience and adjustments in capital structure decisions. Moreover, the analysis uncovers heterogeneous impacts of climate risk on shadow banking, depending on the quality of information disclosure, industry characteristics, and the degree of financing constraints. This research provides new insights into the evolution of corporate financial behavior under climate risk and offers empirical evidence to support firms in optimizing their financial strategies and enhancing their financial risk management capabilities.

1. Introduction

In recent times, with the growing severity of global climate change and the advancement of sustainable development agendas, increasing attention has been paid to how climate-related factors influence economic systems and the decision-making behaviors of individual firms. Climate risk (CRisk) has gradually become a critical factor influencing corporate production, operations, and management decisions. H. H. Huang et al. (2022) document that CRisk deteriorates firms’ financing conditions. Consistently, Ozkan et al. (2023) provide evidence that CRisk adversely affects corporate performance. Firms are subjected to dual pressures from physical and transition risks arising from climate change. Recently, growing academic interest has been directed toward understanding how CRisk affects corporate financial behavior, with particular emphasis on its impact within emerging market economies. Against the backdrop of the dual carbon strategy goals and green transition, how non-financial firms (NFFs) navigate complex external environments, strengthen risk management capabilities, and maintain a robust financial structure has become a critical issue for achieving sustainable development.

From a macroeconomic and financial stability perspective, climate-related risks have increasingly become a source of systemic risk, with growing threats to the stability of the global financial system. As early as 2019, the International Monetary Fund (IMF) included CRisk in its framework for analyzing financial stability, as outlined in its Global Financial Stability Report. Z. Liu et al. (2024) used cross-country panel data analysis to discover that CRisk can impact the stability of the financial system. CRisk has a significant impact on financial markets and investor behavior (J. Yang & Geng, 2025) and, to a certain extent, increases the predictability of stock market volatility (Zhou & Ma, 2025). X. Xu et al. (2025) demonstrate the existence of a non-linear relationship between climate-related risks and the market performance of energy-related assets. Meanwhile, Benkraiem et al. (2025) examine the implications of transition risk for the corporate bond market. However, existing research still has significant shortcomings in its exploration at the micro-enterprise level. In particular, research on how CRisk affects informal financial behavior at the enterprise level remains relatively scarce.

Shadow banking (SBank) refers to a range of non-bank financial institutions engaged in financial intermediation and operating outside the traditional banking regulatory framework. This system encompasses hedge funds, private equity funds, structured financing instruments, trust plans, and various asset management products. Compared to traditional banks, its core characteristic is reliance on short-term financing to support long-term assets, achieving credit expansion through high leverage and regulatory arbitrage. For NFFs, SBank provides an important alternative financing channel, particularly playing a critical role for companies that struggle to secure funding through traditional credit channels. However, J. Li and Han (2019) find that SBank practices among NFFs tend to be high-risk and lack transparency, and are frequently driven by motives of regulatory arbitrage. The characteristics of these activities, including high leverage, information asymmetry, and risk interconnectedness, can further exacerbate firms’ operational risks.

The development of China’s SBank system is closely related to macroeconomic transformation. As economic restructuring deepens and the multi-tiered development of capital markets progresses, corporate financing needs are showing a significant trend toward diversification. Against this backdrop, the scale of the SBank system has grown exponentially. The latest data indicate that by 2024, the total size of China’s SBank system had exceeded CNY 40 trillion. According to data released by the People’s Bank of China, the total assets of China’s financial institutions reached CNY 495.59 trillion by the end of 2024, highlighting the significant role of the SBank sector within the financial system. NFFs, as core participants in the SBank ecosystem, primarily engage in three typical business models: first, short-term capital allocation through the purchase of bank wealth management products, refers to investment products offered by commercial banks to individual or institutional investors; second, investment in trust plans to achieve relatively high returns; and third, participation in various asset management plans for SBank financing. Scholars consistently identify three primary motivations for corporate engagement in SBank: alleviating financing constraints, improving capital allocation efficiency, and pursuing excess returns (J. Huang et al., 2021; Si et al., 2022; Han et al., 2023; Y. Ma & Hu, 2024).

While SBank provides investment channels or alternative financing channels for NFFs, SBank’s complex business model and lack of transparency significantly increase the vulnerability of the financial system. Owing to regulatory loopholes, SBank activities often involve serious maturity mismatches and risk contagion issues, which not only increase the risk exposure of individual financial institutions, but also create systemic risk through cross-holdings and guarantee chains, ultimately posing a threat to financial stability. Empirical studies demonstrate that uncontrolled SBank expansion erodes the resilience of traditional banks while increasing system-wide risk exposure (Ouyang & Wang, 2022; Pan & Fan, 2024). In particular, NFFs’ overreliance on non-standard financial instruments, whether for financing purposes or profit maximization, may distort funding demand and divert it away from the real economy. This dynamic ultimately undermines the efficiency of financial resource allocation. Within this context, CRisk may affect NFFs’ SBank activities by altering firms’ financial conditions, including liquidity, investment opportunities, and financial leverage decisions. Therefore, this study examines how CRisk, as an emerging and critical factor, shapes NFFs’ engagement in SBank at the micro level, providing insights into the mechanisms through which environmental risks influence corporate financial behavior.

This study advances existing research in the following ways: Firstly, it breaks through the limitations of traditional research, which focuses on internal governance and market factors, and confirms the inhibitory effect of external environmental factors on corporate SBank activities, thereby expanding research on the impact of environmental uncertainty on corporate financial decisions. Secondly, the study reveals two key transmission channels: weakened corporate resilience (CorpRes) and capital structure (CapStruc) adjustment. These channels help deepen our understanding of the interaction between environmental factors and corporate finance. Finally, the analytical framework for CRisk and corporate informal financing behavior, which is constructed based on emerging market characteristics, provides a new perspective for assessing climate-related risks and offers insights for improving climate-related financial risk regulation.

The structure of the paper is organized as follows: Section 2 reviews the key literature within the framework of this study and synthesizes the research hypotheses. Section 3 outlines the research methodology, including sample selection and data sources. Section 4 presents the empirical analysis, encompassing robustness tests, mechanism tests, and heterogeneity analyses. Section 5 offers further discussion. It extends and deepens the analysis informed by the empirical results. Finally, Section 6 concludes the paper by summarizing the main findings, offering suggestions for future research, and discussing the policy implications.

2. Literature Review

2.1. CRisk and Corporate Financial Decision-Making

Regarding the classification of CRisk, existing research generally identifies two main types: physical risks and transition risks. Physical risks primarily refer to natural disasters linked to climate change, including extreme weather events such as floods, droughts, typhoons, and other catastrophic environmental occurrences, as well as major climate change events and risks such as global warming, glacier melting, and sea-level rise. Physical risks may lead to production disruptions, supply chain disruptions, and asset damage, thereby impacting production systems and an enterprise’s ability of a going concern. In contrast, transition risk primarily arises from technological innovation, regulatory shifts, and market dynamics associated with the low-carbon transition. With the advancement of the carbon peak and carbon neutrality strategy, regulatory authorities are increasingly tightening their supervision of key emitting industries. The implementation of systems such as carbon quota allocation and emissions monitoring has significantly enhanced compliance costs for enterprises, forcing them to accelerate their green transition process. These impacts not only threaten the short-term operations of enterprises but also pose systemic challenges to their long-term development.

When it comes to investment decisions, climate change is becoming a key factor that enterprises cannot ignore. Existing research indicates that rising CRisk has significantly altered corporate investment behavior patterns: on the one hand, enterprises increase their cash reserves out of precautionary motives to cope with potential climate shocks (Javadi et al., 2023; R. Ma et al., 2024); on the other hand, increased exposure to CRisk reduces investment efficiency and damps corporate innovation activities (W. Xu et al., 2024). Arian and Naeem (2025), through a cross-country comparative study, further demonstrate that corporates located in high CRisk regions tend to pursue more conservative investment strategies to cope with financial and operational uncertainties. Such conservative tendencies reflect a prudent assessment of CRisk in corporate strategic decision-making.

From a CapStruc perspective, the impact of CRisk is particularly significant. In terms of debt structure, Ginglinger and Moreau (2023) and Jiang et al. (2025) confirm that corporates respond to CRisk by optimizing their debt maturity structure, which manifests itself in lower leverage ratios and higher proportions of Long-term liabilities. Empirical studies by Kling et al. (2021) and He et al. (2025) demonstrate that CRisk premiums are reflected in debt financing costs, driving up credit spreads on bonds issued by corporates with high CRisk. Equity financing markets are also deeply affected by CRisk. Yue et al. (2024) and Cepni et al. (2024) confirm that investors demand higher expected returns from companies with high CRisk exposure, which directly enhances equity financing costs. These changes highlight the significant impact of CRisk on corporate CapStruc and financing costs.

In terms of financing innovation, green financial instruments have demonstrated unique advantages. Under the guidance of green finance policies, some corporates have begun to utilize innovative tools such as green bonds for financing (Guesmi et al., 2025), a transition that both meets funding needs and improves environmental performance. Moreover, Chang et al. (2024) empirically document that CRisk also drives firms to adjust their dividend policies, manifesting in reduced cash dividend ratios and alterations in payout methods. This evidence collectively shows that CRisk affects corporate financial decisions through multiple channels, forming a complete chain of transmission from CapStruc to financing costs to profit distribution policies.

Regarding market pricing mechanisms, the impact of CRisk is multifaceted. The pricing mechanism for CRisk in capital markets has become a hot topic in financial research. B. Lin and Wu (2023) analyze and demonstrate that firms’ voluntary disclosure of climate-related information reduces information asymmetry and thereby mitigates the risk of stock price crashes. Empirical data show that the exposure to CRisk is significantly negatively correlated with corporate valuation, with high-CRisk firms generally facing valuation discounts (Q. Li et al., 2024; Berkman et al., 2024). This discount effect reflects the market’s rational pricing of climate transition risks. It is worth noting that the asset revaluation process triggered by CRisk exacerbates financial market volatility. Campiglio et al. (2023) posit through their research that CRisk triggers asset repricing, thereby inducing financial market instability. This evidence collectively suggests that climate factors have deeply influenced the pricing system of capital markets and affect the efficiency of financial resource allocation through channels such as valuation adjustments, risk premiums, and volatility transmission, consequently affecting financial stability.

2.2. SBank by NFFs

In terms of policy impact, existing research suggests a close link between the regulatory environment and corporate SBank activities. Research shows that when monetary policy tightens or financial regulation is strengthened, the scale of corporate financing through the SBank system tends to expand accordingly, a pattern particularly evident in emerging markets with relatively less-developed financial systems (K. Chen et al., 2018; L. Yang et al., 2019). The empirical analysis by G. Lin and Ouyang (2024) points out that the strengthening of China’s macroprudential regulatory policies has significantly enhanced the probability of enterprises financing through SBank channels. Macroprudential policies with distinctive national characteristics include the Differentiated Housing Credit Policy, the Capital Surcharge for Systemically Important Financial Institutions (SIFI), and the Macroprudential Regulation of SBank, and other similar measures. Similarly, the study by H. Chen and Lin (2024) confirms that local fiscal difficulties will prompt local enterprises to actively engage in SBank activities.

However, in terms of the effectiveness of policy interventions, a series of new discoveries have been made in recent years. Guo et al. (2023) argue that small and medium-sized enterprise credit support programs can effectively alleviate financing constraints, thereby reducing firms’ reliance on SBank. Strengthened environmental regulation has also been shown to have a restraining effect, with L. Feng et al. (2023) confirming that strict environmental regulations significantly curb the expansion of SBank activities. In particular, Gu et al. (2024) demonstrate that green credit policies effectively suppress related commission loan practices among heavily polluting enterprises. X. Huang et al. (2024) reveal that digital tax enforcement curbs SBank activities by improving transparency and governance efficiency. Furthermore, J. Zhang et al. (2025) empirically demonstrate that the withdrawal of government implicit guarantees significantly reduces private firms’ involvement in SBank activities. These findings provide important insights into the relationship between regulatory policies and the evolution of SBank.

From the financing supply perspective, structural deficiencies in financial markets are a key driver of corporate participation in SBank, especially in emerging markets where credit resource allocation remains uneven. Allen et al. (2019) identify that, when facing financing constraints, firms tend to obtain funds through informal financial channels, a phenomenon that is particularly pronounced during credit tightening cycles. Han and Li (2020) further reveal that there is a significant positive correlation between the degree of imbalance in financial resource allocation and the scale of SBank activities by NFFs. Bai et al. (2020) also support this view, confirming that credit mismatches do indeed expand the scale of entrusted loans between enterprises. It is worth noting that different financial reform measures have different effects on SBank activities. Jiang et al. (2023) argue that enhancing competition in the banking market can effectively curb the expansion of SBank activities by NFFs. Meanwhile, recent evidence from Si and Liu (2025) shows that advancing capital market liberalization significantly reduces firms’ reliance on SBank financing by improving credit allocation efficiency. These observations highlight the positive role of financial market reforms in curbing SBank activities.

In terms of the impact of technological innovation, the development of financial technology is reshaping the characteristics of corporate SBank activities. Q. Zhang et al. (2023) detect that the widespread application of fintech has opened up diversified financing channels for firms and expanded the operational boundaries of SBank. The empirical analysis by X. Zhao and Yao (2024) further reveals that there is a positive correlation between the degree of enterprise digitalization and their participation in SBank activities, indicating that the depth of technology application has a significant impact on enterprises’ financing choices. However, the research also discovered another effect of technological innovation. Y. Feng et al. (2024) argue that fintech can effectively reduce firms’ reliance on informal financial channels by enhancing information transparency and optimizing capital allocation. In the field of green finance, C. Zhao et al. (2024) confirm that improvements in supply chain Environmental, Social, and Governance (ESG) performance significantly suppress corporate SBank activities. Existing research findings indicate that the impact of fintech on corporate SBank behavior demonstrates a clear duality: while technological innovation creates new development opportunities for SBank activities, technology-driven improvements in the financial system may simultaneously reduce corporate participation in such activities.

Besides external factors, the intrinsic drivers behind corporate participation in SBank activities are also worthy of attention, with corporate governance structures and management characteristics playing a key role. Ren and Shao (2022) find that companies with higher levels of equity structure diversification, particularly those that introduce non-state-owned shareholders, exhibit significantly higher levels of SBank activity. C. Yang and Shen (2022) approach the issue from the perspective of executive background and find that management teams with financial industry experience are more inclined to engage in such activities. From a behavioral finance standpoint, Shi and Zhi (2023) confirm that managerial overconfidence substantially increases the extent to which companies engage in SBank activities. Yan et al. (2024) further propose from a cultural dimension that Confucian traditional values may encourage companies to engage in informal financial activities. Notably, effective external governance mechanisms can play a balancing role. H. Liu et al. (2023) emphasize that positive interaction between institutional investors and firms can reduce information asymmetry and effectively suppress the scale of SBank operations. These research results together demonstrate an important principle: a corporation’s financial decisions are shaped not only by the external market environment but also closely connected to its internal governance features and the behavioral tendencies of its management team.

Existing research has systematically explored the motives and mechanisms behind NFFs’ participation in SBank activities, primarily analyzing these from the perspectives of financial regulation, market structure, and corporate governance. Scholars have employed diverse perspectives and empirical methods to provide a crucial academic foundation for understanding informal financial activities at the firm level. However, existing research has paid insufficient attention to CRisk as a significant external shock factor, particularly in emerging market contexts, where SBank plays a more prominent role in corporate financing due to institutional and financial constraints, and the mechanisms through which CRisk influences firms’ financing decisions remain poorly understood. Given this research gap, this paper focuses on examining the inhibitory effect of CRisk on corporate SBank activities and its transmission mechanisms. As CRisk intensifies, the external uncertainty faced by firms significantly increases. Specifically, the rise in CRisk may prompt firms to reduce their reliance on the SBank system to avoid potential risks.

From a macro-theoretical perspective on CapStruc, firms typically face a trade-off between potential benefits and financial costs when making financing decisions (Modigliani & Miller, 1958). The Trade-off Theory suggests that moderate levels of debt allow firms to benefit from tax shields and financing convenience, thereby reducing overall capital costs; however, excessive leverage exposes firms to significant financial distress costs and bankruptcy risk. Within this theoretical framework, a firm’s participation in SBank activities is influenced not only by internal financing needs and CapStruc optimization but also by external environmental uncertainties. Specifically, CRisk, as a salient external shock, increases the uncertainty of future cash flows and elevates potential operational and financing costs. Consequently, firms may adopt a more cautious approach in selecting financing instruments, thereby reducing their reliance on high-risk, less liquid SBank channels. Based on a systematic review of the theoretical framework and existing research gaps in the literature, this paper proposes the following core research hypotheses:

Hypothesis 1.

Rising CRisk Dampens SBank activities in NFFs.

2.3. CRisk, CorpRes and CapStruc

CorpRes tends to be defined as an enterprise’s ability to respond quickly, operate steadily and develop sustainably in the face of external shocks such as economic fluctuations, technological change, natural disasters, and social and environmental changes. Resilience is increasingly seen as a key indicator of an enterprise’s risk tolerance and long-term competitiveness. Extreme weather events can directly disrupt an enterprise’s production system, driving up operational uncertainty. At the same time, in order to comply with increasingly stringent environmental regulations and carbon emission control mechanisms, companies need to invest heavily in technological innovation, energy conservation, emission reduction, and environmental management, thereby increasing their financial burden and weakening their financial stability. The mechanism by which climate shocks affect corporate efficiency has been verified by a number of empirical studies. Song et al. (2023) and L. Chen et al. (2023) provide evidence that CRisk leads to a significant decline in total factor productivity, with this efficiency loss directly reflected in the sustained weakening of corporate profitability. As CRisk intensifies, increased external environmental uncertainty interacts with declining internal operational efficiency, weakening CorpRes and sustainable development potential, leading to a decline in corporate risk management capabilities, and ultimately eroding the foundation for corporate engagement in SBank activities.

In terms of financing environments, financial institutions’ enhanced ability to identify CRisk significantly impacts credit allocation effects. Businesses with higher climate vulnerability not only face stricter pre-loan reviews but also experience systematically rising financing costs due to risk premiums. This dual pressure forces enterprises to optimize their CapStruc: on the one hand, they reduce their dependence on Short-term liabilities to avoid refinancing risks; on the other hand, they increase the proportion of long-term stable financing instruments. Tran et al. (2024) analyze corporate financial data following the Paris Agreement and find that climate vulnerability exhibits a significant negative correlation with debt ratios, while showing a positive association with firms’ preference for long-term capital instruments. This finding confirms the structural impact of CRisk on corporate financing decisions.

The transformation of corporate CapStruc is showing a clear trend toward long-termization. In response to CRisk, corporations are gradually reducing their Short-term liabilities ratios and rebalancing their financing term structures through equity financing, long-term loans, green bonds and other tools, adopting management decisions to increase the proportion of Long-term liabilities. Among these, green bonds, which combine cost advantages with policy support, have become an important tool for driving corporate low-carbon transformation. However, it is worth noting that such medium- to long-term financing tools often come with stricter disclosure requirements and restrictions on fund usage, which to a certain extent restricts the financial flexibility of enterprises, thereby reducing the space for enterprises to engage in SBank activities.

From the macro-theoretical perspective of CapStruc, firms need to carefully consider potential benefits alongside the risks they may incur when making financing decisions. According to the trade-off theory (Modigliani & Miller, 1958), excessively high leverage significantly increases the likelihood of financial distress and bankruptcy. In the context of rising CRisk, the external uncertainty faced by firms is further amplified, making short-term, high-leverage financing instruments particularly exposed to risk. Therefore, when adjusting their CapStruc, firms must consider not only the traditional trade-off between tax benefits and bankruptcy costs but also the impact of CRisk on financial stability, operational resilience, and long-term sustainability.

Organizational resilience theory emphasizes that a firm’s ability to respond quickly, operate steadily, and maintain sustainable performance in the face of external shocks, such as extreme climate events or changes in environmental policy, is a key determinant of effective risk management. Following disruptive events, CorpRes is reflected in a firm’s capacity and speed to restore normal performance levels (Van Der Vegt et al., 2015). Firms with lower CorpRes tend to reduce the use of high-risk or highly illiquid financing instruments, including SBank activities, under conditions of high uncertainty in order to safeguard operational stability.

Integrating the trade-off theory and the CorpRes perspective provides a theoretical framework for understanding how firms adjust their financing behavior under CRisk. Firms optimize their CapStruc and strengthen CorpRes to cope with external shocks, which in turn indirectly affects their participation in SBank activities. Based on this macro-theoretical analysis, Hypotheses 2 and 3 are proposed to investigate how CRisk affects SBank activities through its impact on CorpRes and CapStruc.

Hypothesis 2.

CRisk dampens SBank activities in NFFs by weakening their CorpRes.

Hypothesis 3.

CRisk dampens SBank activities in NFFs by influencing their CapStruc.

3. Materials and Methods

3.1. Variable Definition

This study uses SBank as the dependent variable. In constructing the SBank indicator, this study draws on the methodology proposed by J. Li and Han (2019). SBank activities tend to fall into two broad categories: the first includes entrusted loans, entrusted wealth management, and private lending; the second category includes wealth management products, trust products, structured deposits, and asset management plans. We will add 1 to the sum of entrusted loans, entrusted wealth management, private lending, and quasi-financial asset management businesses, then take the outcomes of logarithmic normalization as the SBank indicator.

CRisk is the core independent variable. The higher the CRisk, the greater the climate-related risks faced by the enterprise. This paper uses the Management Climate Concern Index developed by Lei et al. (2023) as a measure of corporate CRisk. This index captures the extent to which corporate management addresses climate-related issues in their reports and the intensity of their statements. It not only reflects the level of CRisk perception, climate action, and disclosure by companies but also reveals, to some extent, their attitudes toward the regulatory, market, and reputational impacts of climate change.

Based on the relevance of the research topic, we carefully selected control variables at the firm level that may influence SBank, aiming to comprehensively cover all potential influencing factors. The reasonable configuration of control variables lays a solid foundation for a comprehensive analysis of the complex relationship between CRisk and SBank. The control variables include (1) LogSize, (2) CurrRatio, (3) TobinQ, (4) Tangible, (5) StateOwn, (6) MgmtSh, (7) BoardSize, and (8) IndepDir. The definitions of each variable are detailed in

Table 1. Overview of variables and their definitions.

Variables	Abbreviation	Definitions	Source
Shadow banking	SBank	SBank is measured as the natural logarithm of the sum of entrusted loans, entrusted wealth management, private lending, and quasi-financial asset management activities, plus 1	Adapted from J. Li and Han (2019)
Climate risk	CRisk	Utilize the Managerial Climate Attention Index	Adopted from Lei et al. (2023)
Log of Total Assets	LogSize	The firm’s total assets after applying a natural logarithmic transformation	CSMAR database
Current Ratio	CurrRatio	The firm’s current assets scaled by its current liabilities	CSMAR database
Tobin’s Q	TobinQ	Market capitalization divided by total asset value	CSMAR database
Tangible Asset Ratio	Tangible	Tangible assets scaled by total assets	CSMAR database
Nature of Ownership	StateOwn	Classify enterprises as state-owned enterprises or non-state-owned enterprises according to the nature of their ownership	CSMAR database
Management Shareholding	MgmtSh	The total equity held by the executive team, calculated as their combined shares relative to total shares issued	CSMAR database
Board Size	BoardSize	Total count of board members in the firm	CSMAR database
Independent Director Ratio	IndepDir	Share of independent directors within the company’s board	CSMAR database

.

3.2. Estimation Model

In order to examine the effect of CRisk on SBank activities within NFFs, we compiled a longitudinal dataset of publicly listed non-financial companies in China. An empirical analysis was then carried out using a fixed effects regression model, which helps control for the potential influence of confounding variables. The model is presented as follows:

SBank_i,t = γ₀ + γ₁CRisk_i,t + γ₂LogSize_i,t + γ₃CurrRatio_i,t + γ₄TobinQ_i,t + γ₅Tangible_i,t +
γ₆StateOwn_i,t + γ₇MgmtSh_i,t + γ₈BoardSize_i,t + γ₉IndepDir_i,t + φ_i + λ_t + ε_i,t

(1)

Among these, SBank_i,t is the dependent variable, representing the shadow banking degree of company i in the t-th year; CRisk_i,t is the core independent variable, representing the climate risk level of company i in year t. To control for other factors at the firm level that may influence SBank behavior, the model includes eight control variables, such as LogSize, CurrRatio, TobinQ, and Tangible. To enhance the accuracy of the estimates, we introduce φ_i and λ_t to represent firm and year fixed effects, respectively, to control for unchanging individual characteristics and systematic disturbances at the year level. ε_i,t represents the standard error term.

3.3. Sample Selection and Data Source

We selected NFFs listed on China’s A-share market as our sample, with a time span covering 2007–2023. To ensure that the samples were more targeted and representative, we processed the original samples according to uniform standards: the sample excludes companies from the financial and real estate sectors; it also excludes all Special Treatment (ST) companies and delisted companies. The sample scope covers different regions and industries, and the rich sample size provides a solid and reliable foundation for the study. Unless otherwise specified, all data and indicators are sourced from the China Stock Market & Accounting Research (CSMAR) database. This study employs Stata 18 software for data analysis. To mitigate the impact of outliers on the results, all continuous variables were winsorized at the 1st and 99th percentiles.

3.4. Descriptive Statistics

Table 2. Descriptive statistics for variables.

Variable	Obs.	Mean	Std. Dev.	Min	P25	Median	P75	Max
SBank	44,013	18.5229	2.3116	12.8515	16.8174	18.7036	20.2980	23.4405
CRisk	41,540	12.7609	10.3901	1.8247	5.9699	9.3715	15.6119	55.0636
LogSize	45,387	22.0743	1.2843	19.6455	21.1546	21.8874	22.7959	26.0797
CurrRatio	45,389	2.6562	2.8728	0.2892	1.1346	1.6981	2.9447	18.0746
TobinQ	44,764	2.0376	1.2946	0.8546	1.2575	1.6208	2.3020	8.5908
Tangible	45,387	0.9310	0.0814	0.5537	0.9193	0.9580	0.9790	1.0000
StateOwn	44,460	0.3637	0.4811	0.0000	0.0000	0.0000	1.0000	1.0000
MgmtSh	43,925	14.3627	20.1638	0.0000	0.0011	0.9259	27.2834	69.1580
BoardSize	45,347	8.5262	1.6990	5.0000	7.0000	9.0000	9.0000	15.0000
IndepDir	45,344	37.5020	5.3247	30.0000	33.3300	35.7100	42.8600	57.1400

Note: This table presents the descriptive statistics of all variables. Definitions of the variables are given in Table 1.

presents the summary statistics for the variables, providing insights into the fundamental characteristics and distribution of the sample. Specifically, the mean value of SBank is 18.5229, the standard deviation is 2.3116, and the median is 18.7036. The high average value of the SBank indicator indicates a generally strong level of engagement in SBank activities among the sampled firms. This may reflect a tendency among firms to utilize non-traditional financial channels for capital management when facing financing constraints or liquidity pressure. While such practices may enhance financial flexibility to some extent, they could also increase financial uncertainty and potential risk exposure under escalating CRisk. Notably, there are significant differences in this indicator among enterprises within the sample, which may reflect differences in liquidity management capabilities, risk tolerance levels, and financing constraints among enterprises. The average value of CRisk is 12.7609, with a standard deviation of 10.3901 and a median of 9.3715, suggesting considerable variability in the CRisk exposure among companies. This variation could be influenced by factors such as the carbon emission intensity of the company’s industry, the stringency of environmental regulations, and the company’s governance capacity.

4. Empirical Results

4.1. Baseline Regression

Table 3. Results of the baseline regression model.

Variables	SBank	SBank
	(1)	(2)
CRisk	−0.0089 ***	−0.0096 ***
	[0.0010]	[0.0026]
LogSize	1.0837 ***	0.9972 ***
	[0.0098]	[0.0360]
CurrRatio	0.0812 ***	0.0201 **
	[0.0041]	[0.0085]
TobinQ	0.0649 ***	−0.0098
	[0.0087]	[0.0148]
Tangible	−0.2566 **	0.5265 **
	[0.1245]	[0.2486]
StateOwn	−0.4944 ***	0.1254
	[0.0254]	[0.0796]
MgmtSh	0.0084 ***	−0.0102 ***
	[0.0006]	[0.0016]
BoardSize	−0.0835 ***	0.0211
	[0.0074]	[0.0150]
IndepDir	0.0003	0.0013
	[0.0022]	[0.0036]
Constant	−4.6621 ***	−4.0495 ***
	[0.2679]	[0.9068]
Firm	No	Yes
Year	No	Yes
Observations	37,827	37,652
Adj R-squared	0.2643	0.6099

Note: Numbers in parentheses represent robust standard errors. Coefficients denoted by ** and *** are statistically significant at the 5% and 1% levels, respectively.

reports the benchmark regression results. In column (1), firm and year fixed effects are not included, while column (2) further controls for these fixed effects to account for unobserved firm heterogeneity and time variation. The results indicate that CRisk has a statistically significant negative effect on SBank at the 1% significance level. Specifically, firms facing higher levels of CRisk engage less in SBank activities, and this relationship remains robust after controlling for fixed effects. This finding is consistent with Hypothesis 1 proposed in the literature review, which is grounded in trade-off theory, suggesting that firms reduce high-risk financing activities under higher CRisk to mitigate financial distress. Detailed results are presented in columns (1) and (2) of Table 3.

These benchmark regression results primarily reveal a statistical association between CRisk and SBank, rather than a causal effect. Although the inclusion of fixed effects helps control for unobserved firm heterogeneity and time variation, potential endogeneity issues may still exist, such as reverse causality or omitted variable bias. To further examine the causal effect of CRisk on SBank and verify the robustness of the results, we conduct instrumental variable (IV) estimations in Section 4.3.

4.2. Robustness Check

To further test the robustness of the benchmark regression results, this paper follows the method of Du et al. (2023) to calculate the corporate climate risk index (CIndex). Subsequently, we replaced the independent variable CRisk with CIndex and conducted a regression analysis between CIndex and SBank. As shown in column (1) of

Table 4. Robustness check results.

Variables	SBank	Sbank	SBank	SBank	SBank
	(1)	(2)	(3)	(4)	(5)
CIndex	−0.4768 ***
	[0.1642]
CRisk		−0.0099 ***	−0.0082 ***	−0.0047 **	−0.0100 ***
		[0.0026]	[0.0027]	[0.0019]	[0.0026]
SBank_1				0.3997 ***
				[0.0081]
Finance					−0.2449 ***
					[0.0555]
LogSize	1.0011 ***	1.0100 ***	0.9980 ***	0.6580 ***	1.0018 ***
	[0.0350]	[0.0356]	[0.0370]	[0.0272]	[0.0359]
CurrRatio	0.0151 *	0.0084	0.0098	0.0605 ***	0.0194 **
	[0.0080]	[0.0084]	[0.0090]	[0.0076]	[0.0085]
TobinQ	−0.0030	−0.0042	−0.0055	−0.0314 ***	−0.0109
	[0.0142]	[0.0147]	[0.0163]	[0.0116]	[0.0147]
Tangible	0.4828 **	0.5347 **	0.6284 **	0.4923 ***	0.5704 **
	[0.2443]	[0.2462]	[0.2656]	[0.1874]	[0.2486]
StateOwn	0.1076	0.1262	0.1044	0.0616	0.1315 *
	[0.0790]	[0.0784]	[0.0812]	[0.0590]	[0.0798]
MgmtSh	−0.0095 ***	−0.0102 ***	−0.0103 ***	−0.0052 ***	−0.0104 ***
	[0.0016]	[0.0016]	[0.0017]	[0.0013]	[0.0016]
BoardSize	0.0240*	0.0153	0.0215	0.0143	0.0212
	[0.0144]	[0.0146]	[0.0154]	[0.0110]	[0.0149]
IndepDir	0.0009	−0.0001	0.0021	0.0004	0.0014
	[0.0035]	[0.0035]	[0.0037]	[0.0027]	[0.0036]
Constant	−4.1216 ***	−4.3371 ***	−4.2007 ***	−3.9500 ***	−3.2595 ***
	[0.8787]	[0.8917]	[0.9308]	[0.6436]	[0.9259]
Firm	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes
Observations	40,970	37,652	33,812	33,417	37,648
Adj R-squared	0.6212	0.6033	0.6176	0.6813	0.6103

Note: Numbers in parentheses represent robust standard errors. Coefficients denoted by *, **, and *** are statistically significant at the 10%, 5%, and 1% levels, respectively.

, the regression coefficient of CIndex is negative and significant at the 1% significance level, indicating that the higher the CIndex, the lower the degree of corporate participation in SBank activities. This result indicates that the benchmark hypothesis and conclusions are still supported after replacing the independent variable.

We followed the approach of J. Li and Han (2019) by adding 1 to the sum of entrusted loans, entrusted wealth management, and private lending, then taking the logarithm of the data to standardize it as the Sbank indicator. This newly constructed indicator is then used to replace the original dependent variable SBank. Subsequently, a regression test is conducted between CRisk and Sbank. The results indicate that CRisk and Sbank are significantly negatively correlated at the 1% level. The empirical coefficients suggest that after replacing the dependent variable, the research hypothesis and conclusions remain robust and reliable. See column (2) of Table 4 for details.

In 2014, the People’s Bank of China began to implement a series of monetary policy measures, including interest rate cuts and reductions in the reserve requirement ratio, aimed at boosting economic growth. During this time, the SBank sector also saw significant expansion, particularly with the rapid growth of asset management products, such as trust and wealth management products. By 2015, the scale of China’s SBank had reached an all-time high. The credit supply through non-traditional financial channels gradually surpassed that of traditional bank loans, introducing significant uncertainties to the financial market. Furthermore, the overheating of the real estate market heightened the risk exposure of the SBank sector. To mitigate the interference from major anomalous factors, we excluded the samples from 2014 and 2015 and conducted a regression test on the remaining samples. As evident from column (3) of Table 4, CRisk still significantly inhibits SBank at the 1% level, which supports the benchmark hypothesis and conclusions.

Considering that corporate financial behavior often exhibits persistence and path dependence, this study incorporates the first-order lag of the SBank variable into the regression model to account for the influence of previous-period SBank on current-period behavior. As shown in the regression results in Table 4, the coefficient for CRisk is significantly negative at the 5% level. This suggests that even after controlling for past behavior, the negative relationship between CRisk and SBank persists, thus reinforcing the robustness of the baseline regression. Detailed results can be found in column (4) of Table 4.

The level of financial development at the provincial level not only directly affects an enterprise’s ability to obtain formal financing and the cost thereof, but also indirectly influences the impact of CRisk on an enterprise’s financial behavior. It represents a key external factor affecting the SBank activities of NFFs. Thus, including this variable in the robustness check allows for a more precise assessment of the impact of CRisk on corporate SBank behavior, enhancing both the explanatory power of the model and the reliability of the findings. In this study, we use the ratio of total deposits and loans of financial institutions to GDP as an indicator of provincial financial development (Finance). The empirical results indicate that, even after accounting for Finance, the negative relationship between CRisk and SBank remains significant at the 1% level, confirming that the initial conclusions hold. For the specific coefficients, refer to column (5) of Table 4.

In conclusion, to ensure the robustness of the empirical results and the reliability of the conclusions, this paper conducts multiple robustness tests. Specifically, we systematically test the model specification and data processing by replacing the core independent variables and dependent variables, excluding potential anomalous samples, and introducing regional variables. The results of all robustness tests are consistent with the benchmark regression results, further confirming the basic hypothesis and research conclusions of this paper, thereby enhancing the rigor and explanatory power of the research results.

4.3. Endogeneity Test

To mitigate the potential endogeneity problem, this paper employs the IV to eliminate the bias caused by omitted variables. We follow the approach outlined by Cepni et al. (2024) and use the average CRisk level of other companies within the same industry in the same year as an IV. This IV meets two basic conditions: first, the CRisk of other firms within the industry is correlated with that of the target firm, reflecting the common trends of CRisk shocks; second, this variable indirectly influences SBank behavior by affecting the target firm’s own CRisk, and there is no direct association between the variable and the firm’s SBank activities, thereby satisfying the exogeneity requirement.

In the first stage of the IV regression analysis, various statistical tests confirm the validity of the IV. The LM statistic is 1587.733, and the corresponding p-value of the LM test is 0.000, indicating a strong correlation between the instrument and the model. This satisfies the exogeneity requirement and supports the correlation hypothesis. Additionally, the F statistic is 4598.78, which significantly exceeds the 10% critical value of 16.38, suggesting that the IV does not face weak identification issues. The regression results show a statistically significant positive relationship between the IV and CRisk at the 1% level, confirming the validity of the correlation hypothesis.

In the second stage of the regression analysis, the results show a negative and statistically significant relationship between CRisk and SBank at the 5% level. This suggests that CRisk dampens the SBank activities of NFFs. The results further validated by the IV are consistent with the benchmark model, confirming the robustness of the basic assumptions and conclusions. Detailed results can be found in

Table 5. Instrumental variable estimation results.

Variables	CRisk	SBank
	First Stage	Second Stage
	(1)	(2)
IV	0.8075 ***
	(45.3893)
CRisk		−0.0098 **
		(−1.9747)
LogSize	1.2777 ***	0.9969 ***
	(16.8323)	(44.5678)
CurrRatio	0.0025	0.0206 ***
	(0.1849)	(3.3957)
TobinQ	−0.0081	−0.0114
	(−0.2801)	(−1.0940)
Tangible	0.1249	0.5282 ***
	(0.2286)	(3.3165)
StateOwn	0.6844 ***	0.1252 **
	(3.9415)	(2.3862)
MgmtSh	0.0024	−0.0103 ***
	(0.7128)	(−9.4107)
BoardSize	−0.0370	0.0214 **
	(−1.1449)	(2.2094)
IndepDir	0.0040	0.0011
	(0.5134)	(0.4227)
Kleibergen–Paap rk LM statistic		1587.733 ***
Cragg–Donald Wald F statistic		4598.78 [16.38]
Firm	Yes	Yes
Year	Yes	Yes
Observations	37,582	37,582

Note: Numbers in parentheses represent robust standard errors. Coefficients denoted by ** and *** are statistically significant at the 5% and 1% levels, respectively.

.

4.4. Examination of Mechanisms

In this section, we will explore how CRisk influences the SBank behavior of NFFs from two perspectives: CorpRes and CapStruc. Following the research methodology of Lv et al. (2019), corporate performance growth is measured by the cumulative growth rate of sales revenue over the past three years. At the same time, performance volatility is measured by the standard deviation of monthly stock returns for the current year. These two indicators are then combined using the entropy method and given appropriate weights to calculate the CorpRes score, with a stronger score indicating stronger resilience. In addition, CapStruc is measured by the ratio of the sum of Long-term liabilities and Bonds payable to total assets. Therefore, the following two models were specifically constructed for the mechanism test:

CorpRes_i,t = γ₀ + γ₁CRisk_i,t + γ₂LogSize_i,t + γ₃CurrRatio_i,t + γ₄TobinQ_i,t + γ₅Tangible_i,t +
γ₆StateOwn_i,t + γ₇MgmtSh_i,t + γ₈BoardSize_i,t + γ₉IndepDir_i,t + φ_i + λ_t + ε_i,t

(2)

CapStruc_i,t = γ₀ + γ₁CRisk_i,t + γ₂LogSize_i,t + γ₃CurrRatio_i,t + γ₄TobinQ_i,t + γ₅Tangible_i,t +
γ₆StateOwn_i,t + γ₇MgmtSh_i,t + γ₈BoardSize_i,t + γ₉IndepDir_i,t + φ_i + λ_t + ε_i,t

(3)

In the model, CorpRes_i,t and CapStruc_i,t serve as the dependent variables, denoting the corporate resilience and capital structure of firm i in year t, respectively. CRisk_i,t, on the other hand, represents the climate risk faced by firm i in year t. In line with the methodology of the benchmark regression model, we specify control variables such as LogSize, CurrRatio, and TobinQ. φ_i and λ_t represent the fixed effects specific to the firm and the year, respectively, while ε_i,t denotes the error term.

The mechanism test results presented in

Table 6. Channels analysis: CorpRes and CapStruc.

Variables	SBank	CorpRes	SBank	CapStruc
	(1)	(2)	(3)	(4)
CRisk	−0.0096 ***	−0.0001 ***	−0.0096 ***	0.0010 ***
	[0.0026]	[0.0000]	[0.0026]	[0.0003]
LogSize	0.9972 ***	0.0012 *	0.9972 ***	0.0553 ***
	[0.0360]	[0.0007]	[0.0360]	[0.0029]
CurrRatio	0.0201 **	0.0006 ***	0.0201**	0.0008
	[0.0085]	[0.0001]	[0.0085]	[0.0006]
TobinQ	−0.0098	−0.0068 ***	−0.0098	−0.0010
	[0.0148]	[0.0005]	[0.0148]	[0.0009]
Tangible	0.5265 **	0.0182 ***	0.5265 **	−0.1157 ***
	[0.2486]	[0.0044]	[0.2486]	[0.0225]
StateOwn	0.1254	0.0015	0.1254	0.0039
	[0.0796]	[0.0011]	[0.0796]	[0.0067]
MgmtSh	−0.0102 ***	−0.0001 **	−0.0102 ***	−0.0002 *
	[0.0016]	[0.0000]	[0.0016]	[0.0001]
BoardSize	0.0211	0.0006 **	0.0211	−0.0019
	[0.0150]	[0.0003]	[0.0150]	[0.0013]
IndepDir	0.0013	0.0000	0.0013	−0.0003
	[0.0036]	[0.0001]	[0.0036]	[0.0003]
Constant	−4.0495 ***	0.4327 ***	−4.0495 ***	−0.9926 ***
	[0.9068]	[0.0194]	[0.9068]	[0.0743]
Firm	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes
Observations	37,652	30,055	37,652	38,747
Adj R-squared	0.6099	0.9886	0.6099	0.5663

Note: Numbers in parentheses represent robust standard errors. Coefficients denoted by *, **, and *** are statistically significant at the 10%, 5%, and 1% levels, respectively.

show that an increase in CRisk significantly weakens CorpRes at the 1% significance level, indicating that higher CRisk impairs CorpRes. Furthermore, CRisk is positively and significantly associated with CapStruc, also at the 1% level, suggesting that elevated CRisk induces firms to adjust their CapStruc by increasing the proportion of Long-term liabilities. These empirical results lend strong support to the proposed transmission mechanism.

4.5. Heterogeneity Analysis

The quality of information disclosure is generally closely related to firm size, audit type, and governance structure. In this study, we use the classification of “Big Four” versus “Non-Big Four” audits as a proxy to differentiate firm size and information disclosure quality. The results indicate that for firms audited by Non-Big Four auditors, CRisk has a more pronounced inhibitory effect on SBank activities. Specifically, the coefficient is −0.0101 and is significant at the 1% level (see

Table 7. Heterogeneity test results.

Variables	Big Four Audit		Financing Constraints		Intensity of Production Factors
	Yes	No	Weak	Strong	L-Intensive	T-Intensive	A-Intensive
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
CRisk	−0.0062	−0.0101 ***	−0.0015	−0.0143 ***	−0.0033	−0.0089 ***	−0.0192 ***
	[0.0081]	[0.0027]	[0.0034]	[0.0040]	[0.0051]	[0.0034]	[0.0061]
LogSize	0.9963 ***	0.9964 ***	0.9716 ***	0.9192 ***	1.0116 ***	0.8983 ***	1.0468 ***
	[0.1492]	[0.0369]	[0.0629]	[0.0528]	[0.0684]	[0.0563]	[0.0958]
CurrRatio	0.0629 *	0.0207 **	0.1006 ***	0.0002	0.0432 **	0.0135	0.0192
	[0.0323]	[0.0087]	[0.0167]	[0.0100]	[0.0176]	[0.0112]	[0.0225]
TobinQ	0.0651	−0.0086	−0.0526 **	0.0243	0.0128	−0.0275	−0.0560
	[0.0582]	[0.0153]	[0.0216]	[0.0170]	[0.0313]	[0.0177]	[0.0449]
Tangible	−0.7009	0.5798 **	1.1884 ***	0.4413	0.4553	0.8186 ***	−0.7295
	[1.1577]	[0.2526]	[0.3744]	[0.3096]	[0.4828]	[0.3166]	[0.7169]
StateOwn	−0.1959	0.1391 *	0.1662	0.2707 **	−0.1056	0.2299 *	0.4815 **
	[0.2697]	[0.0820]	[0.1100]	[0.1258]	[0.1349]	[0.1301]	[0.2268]
MgmtSh	−0.0078	−0.0100 ***	−0.0041	−0.0075 ***	−0.0090 ***	−0.0106 ***	−0.0072
	[0.0121]	[0.0016]	[0.0030]	[0.0022]	[0.0033]	[0.0021]	[0.0046]
BoardSize	0.0350	0.0199	−0.0147	0.0397 **	0.0174	0.0284	−0.0493
	[0.0367]	[0.0162]	[0.0248]	[0.0188]	[0.0288]	[0.0238]	[0.0365]
IndepDir	0.0281 ***	−0.0009	−0.0073	0.0060	0.0009	−0.0025	−0.0076
	[0.0082]	[0.0038]	[0.0056]	[0.0048]	[0.0064]	[0.0056]	[0.0098]
Constant	−4.4784	−3.9735 ***	−3.6261 **	−2.7890 **	−4.2178 **	−1.6876	−3.4773
	[3.9451]	[0.9258]	[1.5888]	[1.3016]	[1.7031]	[1.4175]	[2.4448]
Firm	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	2,051	35,541	17,686	19,554	11,175	16,403	6,542
Adj R-squared	0.7859	0.5933	0.5911	0.6937	0.6328	0.6034	0.5681

Note: Numbers in parentheses represent robust standard errors. Coefficients denoted by *, **, and *** are statistically significant at the 10%, 5%, and 1% levels, respectively.

, column (2)). This may be because these firms are more actively engaged in SBank activities, have weaker core business resilience, and exhibit lower information disclosure quality. Consequently, under CRisk shocks, they face greater operational pressure and are compelled to reduce SBank activities. In contrast, firms audited by Big Four auditors are typically larger, possess stronger risk management capabilities, and are less sensitive to CRisk, resulting in more stable SBank activities. In this group, the coefficient is −0.0062 and is not statistically significant (see Table 7, column (1)).

To further investigate the heterogeneous impact of financing constraints on the relationship between CRisk and SBank, we classify firms according to the SA index proposed by Hadlock and Pierce (2010) and divide them at the median into firms with high and low financing constraints. For firms facing high financing constraints, CRisk significantly suppresses SBank activities, with a coefficient of −0.0143, significant at the 1% level (see Table 7, column (4)). In contrast, for firms with low financing constraints, the effect of CRisk on SBank is not significant, with a coefficient of −0.0015 (see Table 7, column (3)). These results suggest that an increase in CRisk intensifies financial institutions’ avoidance of high-risk exposures. For firms already constrained in bank credit, CRisk further compresses available bank funding, thereby exacerbating the reduction in SBank activities.

Considering differences in production factors and industry characteristics, firms in different industries exhibit significant variations in asset and financing structures, which may affect their engagement in SBank activities. Following the approach of Yin et al. (2018), we categorize the sample into labor-intensive, technology-intensive, and capital-intensive firms according to their primary production factors. The analysis shows that CRisk does not have a significant impact on SBank activities for labor-intensive firms, with a coefficient of −0.0033 (see Table 7, column (5)). However, for technology-intensive and capital-intensive firms, CRisk significantly inhibits SBank activities. The coefficients are −0.0089 and −0.0192, respectively, both significant at the 1% level (see Table 7, columns (6) and (7)). This may be attributed to the longer investment cycles and larger funding requirements of technology-intensive and capital-intensive firms, coupled with their better access to formal financial institutions. As a result, when CRisk rises, these firms tend to reduce reliance on informal financing channels such as SBank to lower financing costs.

5. Further Discussion

In the baseline regression results, further examination of the control variables reveals that LogSize is consistently and positively associated with SBank, indicating that larger firms, benefiting from superior resource endowments and better access to information, are more likely to engage in SBank activities. CurrRatio also shows a significant positive effect on SBank, suggesting that firms with stronger liquidity positions are more capable or willing to participate in non-traditional financial activities to enhance capital utilization efficiency. Notably, MgmtSh becomes significantly negatively related to SBank after controlling for fixed effects, implying that as managerial ownership increases and interests between managers and shareholders become more aligned, firms may adopt more prudent financial strategies and reduce reliance on high-risk SBank. In contrast, the effect of StateOwn on SBank becomes insignificant in the extended model, suggesting that after accounting for firm-specific effects, the marginal impact of ownership structure on SBank participation is diminished. Overall, the results for the control variables not only reinforce the robustness of the baseline relationship but also highlight the heterogeneous influence of firm characteristics on SBank decisions.

The mechanism analysis suggests that CRisk influences SBank behavior indirectly by weakening firms’ resilience and prompting adjustments in CapStruc. Empirical results show that an increase in CRisk significantly reduces operational stability and a firm’s ability to withstand shocks, while simultaneously encouraging a shift toward a higher proportion of Long-term liabilities. This indicates a more conservative financing strategy under environmental uncertainty, with firms favoring stable, long-term funding sources. Moreover, the heterogeneity analysis reveals that the impact of CRisk is significantly moderated by the quality of information disclosure and the degree of financing constraints. Specifically, firms with lower transparency or more severe financing constraints experience a stronger dampening effect of CRisk on SBank participation, potentially due to reduced credit access or downgraded creditworthiness, which leads to a passive contraction in SBank channels as CRisk rises. At the industry level, firms in technology-intensive and capital-intensive sectors exhibit greater sensitivity to CRisk, highlighting cross-industry variation in financial adjustment capacity in response to climate-related shocks.

The findings highlight the importance for firms to integrate CRisk into their financial decision-making processes and to develop financial management systems focused on resilience. In the face of increasing uncertainty related to CRisk, it is essential for firms to improve financial flexibility, continuously optimize CapStruc, and address liquidity challenges that may result from mismatches in short-term financing. Furthermore, given that the sensitivity to climate-related shocks varies significantly across industries, policymakers should consider industry-specific characteristics and financing conditions when designing regulatory frameworks. In particular, for sectors characterized by high capital and technology intensity, targeted support through mechanisms such as green financing and government-backed guarantees can play a crucial role in easing credit constraints caused by CRisk. These interventions not only strengthen firms’ ability to manage risks but also contribute to a more sustainable allocation of financial resources, thereby supporting the transition to a low-carbon economy.

To mitigate the inhibitory effect of climate-related risk on firms’ participation in SBank, policymakers should adopt targeted regulatory measures. First, a dynamic monitoring system should be established to track firms’ SBank activities and associated fund flows in real time, with particular focus on industries exposed to high CRisk, in order to prevent potential systemic financial risks. Second, financial stress tests and scenario analyses based on CRisk should be conducted to evaluate the resilience of SBank channels and guide regulatory adjustments. In addition, policies can encourage firms to reduce their reliance on high-risk SBank financing, for example, through the issuance of green bonds, the use of green loans, or government-backed guarantees, ensuring stable funding while limiting systemic exposure. For firms with low transparency or severe financing constraints, regulators should provide differentiated guidance and support to strengthen financial resilience and flexibility. These measures not only help mitigate financial risks but also promote the efficient allocation of resources toward low-carbon and green industries, thereby supporting sustainable economic development.

6. Conclusions

This paper selects non-financial A-share listed companies in China as the research sample. Based on the panel data from 2007 to 2023, it systematically examines the impact of CRisk on SBank behavior of firms. The benchmark analysis indicates that an increase in CRisk significantly reduces the extent to which firms engage in SBank activities. This finding is consistent with the predictions of the Trade-off Theory (Modigliani & Miller, 1958), which posits that firms balance the benefits of leverage against financial distress costs. As CRisk amplifies uncertainty and potential distress costs, firms tend to adopt more conservative financing behavior, thereby reducing reliance on high-risk SBank channels. After implementing several techniques, including variable substitution, sample screening, and the inclusion of regional factors, the main result remains highly consistent. Furthermore, this study utilizes an IV approach to address potential endogeneity concerns, further confirming the causal relationship between CRisk and SBank behavior. This result suggests that climate shocks, as a form of systemic risk, have significantly permeated corporate financing behaviors and reshaped corporate risk perceptions toward informal financial channels.

Further mechanism analysis revealed two main channels through which CRisk affects corporate SBank behavior. First, CRisk weakens CorpRes, making corporate development less flexible and thereby limiting the ability and scope of corporate SBank activities. Secondly, CRisk modifies corporate CapStruc, specifically by raising the Long-term liabilities and reducing Short-term liabilities financing, which further dampens SBank behavior. This adjustment reflects firms’ efforts to mitigate financial distress risk in line with the Trade-off Theory, suggesting that under heightened climate uncertainty, firms prioritize financial stability over aggressive financing expansion. In the heterogeneity analysis, we find that the inhibitory effect of CRisk on SBank activities is more pronounced in Non-Big Four audit firms with lower information disclosure quality and in non-labor-intensive industries. For firms facing more severe financing constraints, SBank behaviors are also significantly suppressed as CRisk increases. These findings suggest that firms’ responses to CRisk exhibit significant structural heterogeneity.

Although this paper has conducted a relatively systematic analysis of the impact path of CRisk on SBank of firms, there are still some aspects of the research that deserve further expansion and in-depth exploration. First, in terms of measuring CRisk, future research can attempt to introduce more refined risk indicators. Second, institutional environmental factors can be considered and incorporated into the analytical framework, such as climate policy uncertainty and policies related to the carbon trading market. Third, the research sample can be expanded from a single-country dimension to a multi-country dimension to reveal the heterogeneous responses of firms to CRisk under different institutional backgrounds.

Based on the research findings, several policy implications can be proposed for regulators and firms. First, in promoting the development of green finance, regulatory authorities should enhance real-time monitoring of non-traditional financing channels, such as SBank, with particular attention to capital flows in industries highly exposed to climate-related risks, in order to prevent the accumulation of systemic financial risks. For another, companies should actively improve their CRisk management capabilities. By optimizing their CapStruc and controlling debt levels appropriately, they can bolster their financial resilience and stability, and progressively decrease their dependence on SBank funds. Concurrently, firms should actively develop green financing instruments, such as green bonds, and engage with green investment projects. This approach will enable them to achieve the twin objectives of securing financing and fulfilling environmental responsibilities, thereby enhancing their risk-bearing capacity and long-term value creation potential.

Building on the aforementioned policy implications, this study emphasizes that regulatory authorities should pay particular attention to firms’ financing behaviors in high-CRisk environments, in order to maintain financial market stability and address climate-related risks. At the same time, green finance policies can guide firms to optimize their financing structures, for instance, by supporting the issuance of green bonds, providing government-backed guarantee funds, or offering differentiated credit support. Such measures can enhance firms’ financial resilience and reduce reliance on high-risk SBank channels. Moreover, they contribute to the efficient allocation of financial resources toward low-carbon projects and green industries, thereby promoting sustainable economic development.