From Financialization to Sustainability: The Impact of Climate Risks on Shadow Banking Activities in Non-Financial Firms in China

Abstract

Climate risks are increasingly shaping corporate strategies, raising important considerations for sustainability. This study explores the impact of climate risks on the shadow banking activities of non-financial firms, examining how these risks influence sustainable financial practices. Analyzing a sample of Chinese listed firms from 2010 to 2022, this study finds that climate risks are negatively associated with shadow banking activities, reflecting a shift towards more sustainable financial management. This finding remains robust through various robustness checks and tests for endogeneity. Heterogeneity results indicate that the reduction in shadow activities due to climate risks is more pronounced in firms with higher dependence on external financing and weaker profitability. Mechanism results suggest that climate risks amplify cross-sectional risks for firms, increasing risk sources. Simultaneously, companies enhance their preference for holding cash to address potential risks. The combined effect of these factors leads to a decrease in shadow banking activities among non-financial firms, aligning with a shift towards sustainability. This study provides new insights into understanding how climate risks drive sustainable financial decision-making and enriches the research on the determinants of firm financialization.

1. Introduction

Shadow banking means various financial intermediary services beyond the conventional banking system, typically carried out by non-bank financial institutions. These institutions transform risks related to credit, liquidity, and maturity of financial assets, playing a “pseudo-bank” role. Starting in 2008, shadow banking in China has developed quickly. According to Moody’s data, at the beginning of 2017, China’s shadow banking scale reached a historical peak of 100.4 trillion yuan, accounting for over 80% of the GDP [1], and by the end of 2021, it still amounted to 57 trillion yuan, nearly 50% of the GDP. In addition to regular financial institutions, small loan companies, and financing guarantee companies, an increasing number of non-financial sectors are utilizing diversified funding sources to engage in substantive lending activities, becoming important participants in the shadow credit market [2]. On one hand, shadow banking operations are characterized by cross-market, cross-financial institutions, and cross-asset class transactions, intensifying the contagion and systemic risks in the financial system [3]. On the other hand, the shadow banking activities probably lead to a crowding-out effect on their main businesses, inducing the risk of a downturn in the real economy through a siphoning effect [3] and even lead to stock market crashes and financial crises [4]. Therefore, understanding the motives and mechanisms behind non-financial companies’ shadow banking activities is crucial for curbing the rise of corporate shadow banking and guiding finance back to the real economy.

Corporate governance is a key factor in determining a company’s financial health, and the growing importance of climate risks should not be overlooked. As emphasized by Shahrour [5], companies are crucial in tackling climate change, and their role in managing climate risks is indispensable. Indeed, the disruption caused by climate risks is one of the greatest challenges of this century [6,7]. Climate risks not only have a significant negative effect on business operations [8] but also affect the overall stability of the financial system [9,10]. There is a growing trend among investors and borrowers to incorporate the impact of climate risks into their considerations during corporate financing and assessment [11], and firms themselves are paying more attention to climate risks and their negative effects, treating them as one of the important factors in business decision-making [12]. When considering climate risks’ impact, an important factor to note is highlighted by Roussel [13]. The study indicates that shadow banking, as part of the financial system, is significantly affected by climate factors. Understanding this influence holds significant importance in the formulation of successful risk management tactics and policies. Due to the low short-term returns of the real economy, non-financial firms often opt for high-risk shadow banking activities, which offer relatively high returns and greater secrecy. It is worth considering how the shadow banking activities of companies are affected under the compounded impact of climate risks. The current literature primarily concentrates on the effect of climate change on corporate economic benefits [14], risk exposure [15], and environmental disclosure [16], while the relationship between corporate climate risk perception and shadow banking activities of non-financial firms has yet to be explored. This study addresses this gap by focusing on the climate risk perception of non-financial firms and seeks to answer two key questions: (i) How do climate risks affect the shadow banking activities of non-financial firms? (ii) What are the underlying mechanisms?

To address the questions, we first collected Management Discussion and Analysis (MD&A) texts of all listed companies in China from 2010 to 2022 and conducted text analysis to form a corporate-level climate risk index. Using the index, we investigated the effect of climate risks on the shadow banking activities. The outcomes indicate a negative correlation between climate risks and shadow banking activities, and this conclusion holds even after addressing endogeneity issues. Furthermore, we explored two heterogeneous effects of climate risks on shadow banking activities, focusing on corporate financing constraints and profitability. The findings indicate that the impact of climate risks on shadow banking activities is more pronounced for firms with strong financing constraints and weaker profitability. Further, we found that for firms with weaker profitability, the relationship between climate risks and the credit intermediation of non-financial firms is closer than that with credit chains. This suggests that firms with weaker profitability are more susceptible to the impacts of climate change, leading to a deterioration in their credit status, and thereby increasing their costs and risks as credit intermediaries.

Additionally, we conducted robustness tests to further confirm the results. First, we further decomposed climate risks into physical risks and transition risks. Second, to address data skewness, we used the natural logarithm of shadow banking size as an alternative dependent variable in the model. Third, we changed the estimation method and used different fixed effects to control for unobservable heterogeneity. The results indicate that our conclusions passed all robustness tests.

Next, we conducted two mechanism analyses to explore the potential mechanisms through which climate risks affect the shadow banking activities. Diversified firms, typically involved in multiple business areas, have complex business connections and balance sheet entanglements compared to other firms. Under the impact of climate risks, firms with a higher degree of business diversification are more likely to trigger a more substantial cross-contagion effect with other firms due to the interconnectedness of their operations and balance sheets. Therefore, we expect to find that corporate climate risk perception reduces firms’ willingness to engage in shadow banking activities by increasing cross-risks. Second, climate risks may deteriorate the operational environment of firms, increasing operational costs and cash flow risks. To mitigate external risk shocks, firms may choose to hold more cash, thereby reducing leverage and shadow banking activities. We explored the inhibitory effects of cross-risk and cash holding preferences on the shadow banking activities. Both analyses yielded affirmative results, suggesting that our findings uncover the underlying mechanisms.

Our study makes three contributions. First, in the context of heightened climate risks in China, this study investigates how non-financial firms respond to these risks, thereby extending the existing research on the impact of climate risks [17]. It provides empirical evidence from China on the relationship between extreme climate events and corporate investment and financing behaviors. Second, it enriches the research on the driving factors of shadow banking activities. A series of studies have analyzed these factors from the perspectives of economic policy [18], corporate management [19], and financial technology [20]. Unlike previous research, our study focuses on the exploration of internal factors and, for the first time, investigates the impact of corporate climate risk perception on shadow banking activities, offering references for regulatory authorities to incorporate climate risks into decision-making during financial reform deepening. Lastly, by dissecting the internal mechanisms of firms’ behaviors in the face of climate risk challenges, our study reveals the essence of how climate risks impact corporate investment and financing behaviors. Our findings indicate that the root cause for non-financial firms reducing their shadow banking activities lies in the heightened uncertainty about the future economic environment due to climate risks, providing a comprehensive interpretation of the behavior of non-financial firms in the face of compounded risks.

The rest of this study is structured as follows: Section 2 conducts a literature review and formulates research hypotheses. Section 3 outlines the sample selection and research methodology. Section 4 investigates the impact of climate risks on the shadow banking activities of non-financial firms. Section 5 conducts robustness tests. Section 6 performs a heterogeneous analysis. Section 7 discusses potential mechanisms. Section 8 concludes this study.

2. Literature Review and Hypotheses

2.1. Literature Review

2.1.1. The Impact of Climate Risks

Climate risk means the potential negative consequences for human society and the natural environment resulting from climate change or responses to it [21]. The TCFD report initially pointed out that climate risks can be categorized into two types: physical risks (entity risks) and transition risks. This classification is also adopted in the Basel Committee’s report. Physical risks are direct or indirect losses to natural and man-made assets resulting from immediate extreme weather events (such as droughts, floods, storms, etc.) or long-term changes in climate patterns (such as sea level rise, temperature increases, changes in precipitation, etc.) due to climate change [22]. Transition risks refer to the impact on the revaluation or reallocation of assets and income due to policy, technological, market, and behavioral changes in reaction to climate change. This includes the repricing of high-carbon assets, policy shifts, financial and reputational losses, etc. [23]. These two types of risks interact with each other, jointly constituting the comprehensive impact of climate change [24].

The existing literature has conducted in-depth research on climate risks and their impact on the economy and finance. From the perspective of the financial system, climate-related risks affect macroeconomics and financial stability through channels such as the balance sheets of financial institutions, the supply and demand of bank credit, market liquidity, asset prices, and investment return expectations [25,26,27,28]. Insurance companies, collateral, and financial markets are key conduits in this process [29,30]. Hong points out in his research that the effects of climate change are profoundly significant on the overall economy, as represented by the stock market [10]. Zhang suggests that climate risks drive and exacerbate fluctuations in financial markets from different angles, potentially bringing additional uncertainties and significant risks to financial markets [31]. For firms, climate risks adversely affect their operational activities. Huang suggests that climate risks can reduce a firm’s profitability and cash flow stability, thereby increasing its financial risks [12]. Ozkan, through comparing the performance of firms in multiple countries facing climate risks, finds that these risks can reduce a firm’s profitability and market value [8]. To cope with the challenges posed by climate risks, firms often change their investment and operational activities to maintain profitability [12].

2.1.2. Shadow Banking Activities of Non-Financial Firms

Shadow banking activities refer to financial credit intermediation conducted outside the traditional banking regulatory system, often carried out by non-bank financial institutions. These institutions transform risks related to the credit, liquidity, and maturity of financial assets, playing a “pseudo-banks” role. According to Moody’s data, China’s shadow banking scale has broadened fast since 2008 and peaked in early 2017, accounting for 80% of the domestic GDP [1]. Subsequently, with the gradual improvement of regulatory measures, shadow banking activities have slightly declined but still remain at a relatively high level. The reliance of China’s shadow banking sector is predominantly on conventional banking systems, encompassing wealth management offerings, entrusted loans, asset management agencies, trust businesses, and private loans [32]. Hachem notes that while China’s shadow banking possesses unique characteristics, it fundamentally aligns with the global definition of shadow banking, engaging in credit intermediation with mismatched maturities outside regulatory oversight [33]. Additionally, research by Ahmed suggests that non-financial corporations with robust corporate governance structures and active engagement in corporate social responsibility (CSR) are better positioned to manage financial risks and enhance financial performance [34]. This insight offers a fresh perspective on how non-financial firms leverage shadow banking activities in response to external economic pressures. Effective governance and CSR practices not only bolster firms’ resilience against financial market volatility but also diminish the incentives for excessively investing in high-risk financial products for short-term gains. Non-financial institutions, as important participants in shadow banking, are the focus of existing research regarding their motives for involvement. The motivations for non-financial firms to engage in shadow banking activities can be classified into preventative motives and arbitrage motives [35]. Preventative motives refer to firms using their financing advantages to conduct short-term shadow banking operations to address risks and liquidity issues [36]. Arbitrage motives indicate firms are pursuing high returns by sacrificing the development of their main business and investing in high-risk, low-liquidity projects [37]. The shadow banking activities of non-financial firms in China are more driven by arbitrage motives [3,38]. Simultaneously, they are also influenced by a variety of factors, including financing structure [2], corporate management models [39], and the macroeconomic environment [40].

As previously discussed, climate risks significantly impact government policies and corporate management decisions [41,42]. Addressing climate change-related losses, governments have implemented green credit policies. Through Green Support Factors (GSF), environmentally friendly companies gain increased access to credit funds, while Brown Penalty Factors (BPFs) raise the cost of capital for carbon-intensive businesses [13]. These policies affect the operations of financial institutions, including banks and shadow banks, and have profound impacts on non-financial firms. Specifically, when traditional credit channels are restricted by policy regulations, non-financial companies may turn to shadow banking as an alternative funding source. This choice is driven not only by the flexibility and lower transaction costs offered by shadow banking but also by its ability to circumvent stringent financial regulations, providing a financial “safe harbor” for corporate operations [43]. Moreover, climate risks directly influence the strategic management decisions of non-financial firms, thereby affecting their choices of funding sources. The preference for shadow banking varies among firms, which will be discussed in detail in the subsequent section on research hypotheses. In summary, climate risks influence the scale of shadow banking both indirectly, by impacting corporate funding sources through government policies, and directly, by affecting corporate strategic decisions.

2.2. Research Hypotheses

The relationship between climate risks and shadow banking activities in non-financial firms is dual-faceted.

On one hand, climate risks may reduce shadow banking activities. Climate risks significantly influence government policies and corporate management decisions [41,42], which is particularly critical for non-financial firms. Specifically, climate disasters can directly disrupt business operations, affecting performance [12,44], increasing financing costs [45], damaging supply chains [46], and negatively impacting sales [47]. Additionally, physical and transition risks significantly affect corporate operations and investment decisions. Physical risks may lead to asset damage, revenue decline, and financial deterioration, increasing credit and liquidity risks and reducing firms’ engagement in shadow banking. Transition risks may decrease market competitiveness, erode asset values, and undermine investor confidence, increasing market and reputational risks, thus reducing shadow banking activities.

Furthermore, climate risks indirectly influence shadow banking by exacerbating cross-firm risks and increasing firms’ cash holding preferences. Highly interconnected and diversified business structures amplify the negative impacts of climate risks, potentially leading to higher production costs, lower revenues, asset devaluation, and credit deterioration [48,49], thereby increasing uncertainty and complexity, affecting investment efficiency and returns [50]. In response, firms may prioritize financial stability and independence, reducing reliance on shadow banking. Regarding cash holding preferences, climate risks increase decision-making uncertainty, affecting investment and R&D activities, raising financing costs, and prompting firms to accumulate more cash to mitigate potential risks and shocks [51,52,53]. This strategy is particularly evident in financially constrained firms [54]. Additionally, green credit policies, as a response to climate risks, significantly influence firms’ cash holding preferences, especially by imposing stricter loan requirements on high-polluting firms [55,56], encouraging them to increase cash reserves and reduce dependence on shadow banking.

Drawing from the preceding discussion, this study suggests the subsequent hypothesis:

Hypothesis 1.

Climate risks reduce shadow banking activities in non-financial firms.

On the other hand, climate risks may increase shadow banking activities. First, rising climate risks pose severe threats to the overall stability of the financial industry, prompting the traditional financial system to strengthen regulation, particularly for high-risk firms. This results in increased costs and difficulty in obtaining funds through formal channels, eventually forcing some firms to seek alternative funding sources to buffer risks and meet their financial needs. This increased demand naturally drives firms to rely more on the flexible services provided by shadow banking [57]. Second, in the face of potential losses due to climate risks, firms may seek various investment strategies to hedge against these impacts [58]. Non-traditional financial products offered by shadow banks, such as high-yield bonds and private debt, often attract firms due to their higher returns [18]. Third, in a stringent regulatory environment, particularly in climate risk management, traditional banks may face operational constraints, while the shadow banking system, benefiting from relatively lenient regulations in certain regions, offers firms opportunities to circumvent strict financial oversight [59,60]. This regulatory arbitrage allows non-financial firms to access funds with less regulatory pressure, potentially increasing reliance on shadow banking activities. Finally, under the pressures of climate change, non-financial firms may require more flexible liquidity management tools to respond to rapid market changes and uncertainties. The shadow banking system provides various financial instruments, such as commercial paper and refinancing facilities, which help firms manage their short-term liquidity needs more effectively, thereby increasing their dependence on the shadow banking system [61].

Based on the above, this study presents the following hypothesis:

Hypothesis 2.

Climate risks increase shadow banking activities in non-financial firms.

3. Data and Descriptive Statistics

3.1. Data and Sample Construction

We collected data on all companies listed in China’s A-share market from 2010 to 2020 (4693 in total), excluding financial institutions and ST (Special Treatment) companies. Information on the shadow banking activities of non-financial firms was gathered from the China Securities Market & Accounting Research (CSMAR) database and corporate announcements. Climate risk data were also sourced from the CSMAR database. Information for control variables, including company financial details and market data, was obtained from the CSMAR, RESSET, and Wind databases.

3.2. Corporate-Level Climate Risk Variable

Following the methodology used by Sautner to build a climate risk index for U.S. listed companies [15], this study employed text analysis and word frequency statistics techniques to build a corporate-level climate risk index for Chinese listed firms. In the first step, following Huynh, we used relevant word list and text analysis techniques to construct climate risk keywords [62]. Specifically, we referred to official documents released by the Chinese government, such as “Notice by the State Council of the Action Plan for Carbon Dioxide Peaking Before 2030” and “Notice by the State Council on Issuing the National Plan for Addressing Climate Change,” to select 100 basic words related to climate change and risks to form a climate risk-related word list. In the second step, we used jieba to analyze the MD&A texts of all listed Chinese companies from 2010 to 2022 and counted the frequency of all 100 related words. After eliminating words that appeared fewer than 10 times and manually processing similar words (e.g., solar energy and solar power generation), we ultimately selected 48 high-frequency climate risk-related words to construct a climate risk-related dictionary, as shown in

Table 1. Keywords for climate risk, transition risk, and physical risk.

Transition Risks
Carbon dioxide	Carbon emissions	Carbon peak	Carbon neutrality
Dual carbon	Carbon reduction	Carbon trading	Carbon sink
Low carbon	Decarbonization	Carbon reduction	Zero carbon
Natural gas	Energy saving	Coal power	Green energy
Solar energy	Renewable energy	Coal	Energy storage
Hydropower	Petroleum	Consumption reduction	Energy consumption
Fossil energy	Biomass energy	Energy consumption	Paris Agreement
Wind power	Geothermal energy	Nuclear power	Clean energy
Wind energy	Photovoltaic energy
Physical Risks
Climate	Weather	Drought	Temperature
Snow	Rainfall	Flood	Flooding
Freeze	Drought disaster	Drought conditions	Low temperature
Heavy rain	High temperature

. In the third step, we calculated the annual climate risk index (CC1) for each company by summing the frequency of climate risk-related words appearing in the MD&A texts of each year and dividing it by the total word count in the text. A higher frequency of climate risk-related words indicates greater exposure to climate risks for the company in that year.

3.3. Construction of Shadow Banking of Non-Financial Firms

We developed a metric to assess the shadow banking activities of non-financial firms following Li and Han [63]. These activities occur in two forms: credit intermediation and credit chains. Credit intermediation is quantified by the total of entrusted loans, entrusted wealth management, and private lending, divided by total assets, and is labeled as Credit_Intermediary. In order to assess private loan amounts, “other receivables” serve as a proxy variable for inter-company capital lending [64]. Data on entrusted loans are sourced from announcements made by listed companies [38], whereas information regarding the management of entrusted wealth and private lending is sourced from the CSMAR database.

In the shadow banking system, the credit chain primarily involves firms engaging in credit market investment and financing activities by purchasing financial products such as bank financing, brokerage financing, trust products, and structured deposits. The credit chain is defined as the total of financial products, trust products, structured deposits, and asset management plans, divided by total assets, and labeled as Credit_Chain. Data on financial products are sourced from the “other current assets” section within the “notes to financial statements” in the CSMAR database. Lastly, the overall scale of shadow banking activities for non-financial firms is measured by combining Credit_Intermediary and Credit_Chain and is denoted as Shadow_Banking.

3.4. Control Variables

Building upon existing research [20,63], we also controlled for a series of variables at both the corporate and macro levels, including Leverage, Roa, Size, Lnage, Dual, Tang, Interest, M2, MacroPI, and BankPI. M2, MacroPI, and BankPI represent the impact of the macroeconomic environment on non-financial corporate shadow banking. The Macroeconomic Prosperity Leading Index (MacroPI) and banking industry Prosperity Leading Index (BankPI) reflect business conditions through surveys of entrepreneurs and bankers, respectively.

Table 2. Variable definitions.

Variable	Definition
CC1	Climate risk index, the word frequency of climate risk-related words divided by the total word frequency in MD&A text.
Shadow_Banking (SB)	The shadow banking scale of firms is defined as the sum of credit intermediaries and credit chains divided by total assets.
Credit_Chain (SB_CC)	The shadow banking scale of credit chain. The sum of wealth management products, trust products, structured deposits, and asset management plans divided by total assets.
Credit_Intermediary (SB_CI)	The shadow banking scale of credit intermediaries. The sum of entrusted loans and private lending divided by total assets.
Leverage	Total liabilities divided by total assets.
Roa	Ratio of net income to total equities.
Size	Firm size, measured as the logarithm of assets.
Lnage	Firm age, measured as the natural logarithm of age + 1.
Dual	Whether the chairman and the general manager are the same person (1 for yes, 0 for no).
Tang	The ratio of tangible assets to total assets indicates the proportion of fixed assets in total assets.
Interest	Interest expenses paid by the firm divided by liabilities.
M2	Year-on-year growth rate of M2
MacroPI	Macroeconomic prosperity is the leading index.
BankPI	Banking industry prosperity leading index.

contains detailed descriptions of various variables.

3.5. Descriptive Statistics

Table 3. Descriptive statistics.

Variable	N	Mean	Sd	Min	Max	p50
CC1	38,693	0.2340	0.3858	0.0000	2.1174	0.0821
SB	38,693	0.1108	0.2460	0.0002	1.4046	0.0128
SB_CI	38,693	0.1041	0.2345	0.0002	1.3580	0.0127
SB_CC	38,693	0.0062	0.0160	0.0000	0.0955	0.0000
Leverage	38,693	0.4168	0.2101	0.0502	0.9269	0.4058
Roa	38,693	0.0392	0.0646	−0.2600	0.2109	0.0395
Size	38,693	22.1377	1.3097	19.6587	26.2193	21.9419
Lnage	38,693	1.9838	0.9839	0.0000	3.3673	2.1972
Dual	38,693	0.2242	0.4171	0.0000	1.0000	0.0000
Tang	38,693	0.3452	0.1782	0.0023	0.7810	0.3327
Interest	38,693	14.1386	47.4457	−30.3517	370.0696	2.2059
M2	38,693	0.1144	0.0300	0.0817	0.1973	0.1101
MacroPI	38,693	98.4908	5.4767	91.3400	110.3600	97.9200
BankPI	38,693	70.6949	6.7741	60.5000	87.2000	68.7000

displays the descriptive statistics for our sample. Continuous variables are winsorized at the 1st and 99th percentiles to mitigate the impact of outliers. These data reveal that the mean value of shadow banking is 0.1108, with the mean values for the credit chain and credit intermediary being 0.0062 and 0.1041, respectively. This suggests that credit intermediaries constitute the dominant form of shadow banking activities. The mean climate risk index is around 0.2340, with a standard deviation of 0.3858, indicating substantial variability in the climate risks perceived by companies.

4. Empirical Results

4.1. Baseline Result

The specific model of this study is as follows:

$$Y_{i,t}={\beta }_0+{\beta }_1{Risk}_{i,t}+{\beta }_2{Z}_{i,t}+{\delta }_t+{\phi }_i+{\epsilon }_{i,t}$$

(1)

where $i$ represents the firm and $t$ represents the year, $Y_{i,t}$ is one of the measures for shadow banking activities of non-financial firms: Shadow_Banking, Credit_Chain, and Credit_Intermediary. $Z_{i,t}$ is a vector of control variables that may affect the shadow banking activities of non-financial firms. All variables are defined in detail in Table 2. ${\delta }_t$ are year-fixed effects, and ${\phi }_i$ are firm-fixed effects.

Columns 1–6 of

Table 4. Regression analysis of climate risk on shadow banking activities.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	SB	SB_CC	SB_CI	SB	SB_CC	SB_CI
CC1	−0.0235 ***	−0.0015 ***	−0.0215 ***	−0.0225 ***	−0.0008 **	−0.0212 ***
	(−5.03)	(−4.57)	(−4.88)	(−4.68)	(−2.40)	(−4.69)
Leverage				−0.1197 ***	−0.0099 ***	−0.1091 ***
				(−8.19)	(−9.68)	(−7.92)
Roa				−0.0513 **	−0.0016	−0.0506 **
				(−2.22)	(−0.94)	(−2.30)
Size				−0.0307 ***	−0.0016 ***	−0.0290 ***
				(−8.52)	(−7.01)	(−8.38)
Lnage				−0.0136 ***	−0.0032 ***	−0.0101 ***
				(−3.44)	(−11.19)	(−2.67)
Dual				0.0165 ***	0.0003	0.0158 ***
				(3.32)	(1.05)	(3.33)
Tang				−0.1325 ***	−0.0070 ***	−0.1246 ***
				(−9.23)	(−7.06)	(−9.13)
Interest				0.0000	0.0000	0.0000
				(0.28)	(0.81)	(0.16)
M2				−0.5033 ***	−0.0443 ***	−0.4586 ***
				(−9.58)	(−12.02)	(−9.13)
MacroPI				0.0008 ***	0.0000 ***	0.0008 ***
				(5.95)	(3.96)	(5.93)
BankPI				−0.0044 ***	−0.0003 ***	−0.0040 ***
				(−16.55)	(−18.84)	(−15.85)
Constant	0.1200 ***	0.0049 **	0.1147 ***	1.2388 ***	0.0795 ***	1.1501 ***
	(3.85)	(2.28)	(3.82)	(12.96)	(12.68)	(12.59)
Observations	38,693	38,693	38,693	38,693	38,693	38,693
Adjusted R-squared	0.001	0.002	0.001	0.056	0.068	0.052
Company FE	Yes	Yes	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes	Yes	Yes

Note: T statistics in parentheses. ***, ** indicate significance at the 1%, 5% levels, respectively.

present the empirical findings of the benchmark model. Columns 1–3 show the direct impact of climate risks on shadow banking, credit chain, and credit intermediary, where the variable coefficients are negative and statistically significant, ranging from −0.0235 to −0.0015. That shows when only controlling for fixed effects, climate risks weaken various activities of shadow banking. Column 4 presents the regression results after including control variables, where the coefficient between climate risks and activities of shadow banking remains negative and statistically remarkable at −0.0225. This suggests that as climate risks rise, shadow banking activities tend to decline. The result supports our first hypothesis, which posits a negative relationship between climate risks and shadow banking activities. Additionally, we examined the effect of climate risks on various types of shadow banking activities. Columns 5 and 6 report the outcomes for Credit_Chain and Credit_Intermediary, respectively. The findings reveal that climate risks are negatively related to both credit chain and credit intermediary activities in shadow banking, further confirming Hypothesis 1.

This suggests that with the increase in climate risks, shadow banking activities of non-financial firms tend to decrease. The finding confirms our first hypothesis that there is a negative link between climate risks and shadow banking activities. Additionally, we studied the effect of climate risks on different types of shadow banking activities. Columns 5 and 6 report the results for Credit_Chain and Credit_Intermediary, respectively. The research finds that climate risks are negatively correlated with both the credit chain and credit intermediary activities of shadow banking, further validating Hypothesis 1.

For corporate-level control variables, all coefficients except for Interest are statistically significant across all columns in the regression results. Leverage and Roa primarily reflect firms engaging in shadow banking activities to pursue higher profits, while Size, Lnage, and Tang mainly indicate firms undertaking shadow banking activities for more convenient access to financing channels. These results are consistent with previous empirical research findings [20,65].

In terms of the main macroeconomic variables, the shadow banking activities of non-financial firms show a negative correlation with M2 and BankPI and a positive correlation with MacroPI. This indicates that when M2 increases, non-financial firms can more easily access conventional financing provided by traditional financial institutions (such as banks), thus reducing their reliance on shadow banking. When BankPI rises, banks are more inclined to actively offer loans and financing, which diminishes the likelihood of non-financial firms turning to shadow banking for financing reasons. Conversely, when MacroPI increases, indicating an uptrend in macroeconomic prosperity, firms’ financing needs grow. This includes their demand for shadow banking services to satisfy the requirements for expansion, investment, and operational funds. In summary, consistent with previous research [18], when the overall macroeconomic climate improves, firms’ overall financing needs increase, thereby expanding their demand for shadow banking services.

4.2. Endogeneity Test

Because of potential issues of reverse causality and omitted variable bias, there is an endogeneity problem between climate risks and corporate shadow banking activities. Specifically, climate risks can influence a firm’s willingness to engage in shadow banking activities, while such activities can increase a firm’s financial and operational risks, thus affecting its capacity to respond to climate risks. Therefore, our study adopts Raymond’s approach, using the average climate risk level of other firms within the same industry and year as an instrumental variable. This instrumental variable meets two key criteria: first, the climate risk of other firms in the same industry and year affects the firm’s climate risk index; second, the climate risk of other firms in the same industry and year does not directly influence the shadow banking activities of the firm under analysis.

Table 5. Endogeneity test.

	(1)	(2)	(3)	(4)
Variables	CC1	SB	SB_CC	SB_CI
CC1_IV	0.8086 ***
	(18.52)
CC1		−0.2364 ***	−0.0093 ***	−0.2260 ***
		(−13.95)	(−8.34)	(−14.03)
Leverage	0.0061	−0.1236 ***	−0.0100 ***	−0.1129 ***
	(0.32)	(−11.68)	(−13.50)	(−11.16)
Roa	0.1297 ***	−0.0271	−0.0006	−0.0274
	(4.71)	(−1.31)	(−0.41)	(−1.39)
Size	0.0304 ***	−0.0214 ***	−0.0013 ***	−0.0200 ***
	(5.44)	(−8.70)	(−7.66)	(−8.52)
Lnage	0.0087 *	−0.0046	−0.0028 ***	−0.0014
	(1.75)	(−1.47)	(−12.86)	(−0.47)
Dual	−0.0022	0.0143 ***	0.0003	0.0138 ***
	(−0.48)	(3.77)	(1.01)	(3.79)
Tang	0.0254	−0.1327 ***	−0.0070 ***	−0.1248 ***
	(1.11)	(−12.92)	(−9.89)	(−12.72)
Interest	0.0000	0.0000	0.0000	0.0000
	(0.06)	(0.22)	(0.80)	(0.09)
M2	0.0677	−0.4214 ***	−0.0410 ***	−0.3802 ***
	(1.17)	(−9.06)	(−12.69)	(−8.54)
MacroPI	−0.0003 *	0.0004 **	0.0000 *	0.0004 **
	(−1.73)	(2.17)	(1.81)	(2.18)
BankPI	−0.0003	−0.0048 ***	−0.0003 ***	−0.0044 ***
	(−1.23)	(−22.17)	(−23.37)	(−21.27)
Constant	−0.6208 ***
	(−3.93)
Observations	38,693	38,374	38,374	38,374
Adjusted R-squared	0.136	−0.124	−0.071	−0.129
Kleibergen–Paap RK LM statistic		517.181 ***	517.181 ***	517.181 ***
Cragg–Donald Wald F statistic		2234.466	2234.466	2234.466
		[16.38]	[16.38]	[16.38]
Company FE	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes

Note: ***, **, and * represent significance at the 1%, 5%, and 10% levels, respectively. [] is the critical value at the 10% level of the Stock–Yogo weak identification test.

presents the 2SLS results for the instrumental variable analysis. The Kleibergen–Paap RK LM statistic is remarkable, and the Cragg–Donald Wald F statistic exceeds the critical value for the Stock–Yogo weak instrumental variable identification test at the 10% significance level, rejecting the null hypothesis of a weak instrumental variable. Thus, the instrumental variable used in this study is deemed reasonable and reliable. Columns 2–4 of Table 5 show that climate risks diminish the shadow banking operations of non-financial firms, corroborating the initial findings.

5. Robustness Test

5.1. Examining an Alternative Independent Variable

Risks associated with climate encompass risks related to both physical conditions and transitional situations. We further consider the impact of physical and transition risks on the shadow banking activities. Following the method used in the existing literature [66], we measure physical risks (CC_P1) as the frequency of physical risk-related words divided by the total word frequency, and transition risks (CC_T1) are measured similarly using the frequency of transition risk-related words. We further used the natural logarithm of climate risk-related sentence frequency plus one as a substitute indicator (CC2). The results using these three indicators as substitutes for the core explanatory variable are presented in columns 1–3 of

Table 6. Alternative core independent variables, dependent variables, and estimation methods.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Variables	SB	SB	SB	SB1	SB1	SB1	SB	SB
CC1				−0.3092 ***			−0.0216 ***	−0.0192 ***
				(−5.58)			(−4.57)	(−8.02)
CC2	−0.0062 ***
	(−4.41)
CC_T1		−0.0217 ***			−0.2917 ***
		(−4.41)			(−5.09)
CC_P1			−0.1102 **			−1.9148 ***
			(−2.49)			(−4.53)
Leverage	−0.1210 ***	−0.1196 ***	−0.1195 ***	−0.2276 *	−0.2264 *	−0.2265 *	−0.1229 ***	−0.1761 ***
	(−8.30)	(−8.19)	(−8.19)	(−1.67)	(−1.66)	(−1.67)	(−8.59)	(−24.67)
Roa	−0.0528 **	−0.0516 **	−0.0530 **	−0.8224 ***	−0.8268 ***	−0.8419 ***	−0.0522 **	0.0823 ***
	(−2.29)	(−2.24)	(−2.30)	(−4.00)	(−4.02)	(−4.10)	(−2.26)	(3.86)
Size	−0.0300 ***	−0.0308 ***	−0.0314 ***	0.7857 ***	0.7841 ***	0.7774 ***	−0.0314 ***	−0.0083 ***
	(−8.30)	(−8.54)	(−8.72)	(23.15)	(23.11)	(23.03)	(−8.97)	(−8.79)
Lnage	−0.0126 ***	−0.0137 ***	−0.0141 ***	−0.0897 ***	−0.0918 ***	−0.0946 ***	−0.0132 ***	−0.0119 ***
	(−3.15)	(−3.48)	(−3.58)	(−2.65)	(−2.71)	(−2.80)	(−3.37)	(−7.86)
Dual	0.0161 ***	0.0165 ***	0.0166 ***	0.0807 **	0.0810 **	0.0823 **	0.0162 ***	0.0085 **
	(3.24)	(3.32)	(3.34)	(2.06)	(2.07)	(2.10)	(3.29)	(2.40)
Tang	−0.1323 ***	−0.1324 ***	−0.1326 ***	−0.8596 ***	−0.8577 ***	−0.8620 ***	−0.1269 ***	−0.1378 ***
	(−9.24)	(−9.22)	(−9.24)	(−6.11)	(−6.10)	(−6.14)	(−9.04)	(−21.31)
Interest	0.0000	0.0000	0.0000	0.0003	0.0003	0.0003	0.0000	0.0001 *
	(0.25)	(0.28)	(0.28)	(1.18)	(1.19)	(1.17)	(0.33)	(1.75)
M2	−0.4961 ***	−0.5044 ***	−0.5081 ***	−4.2202 ***	−4.2375 ***	−4.2729 ***	−0.4970 ***	−0.6039 ***
	(−9.41)	(−9.60)	(−9.67)	(−9.52)	(−9.56)	(−9.65)	(−9.47)	(−6.07)
MacroPI	0.0008 ***	0.0008 ***	0.0009 ***	0.0056 ***	0.0057 ***	0.0058 ***	0.0008 ***	0.0012 ***
	(5.45)	(6.01)	(6.14)	(5.15)	(5.26)	(5.40)	(6.02)	(4.17)
BankPI	−0.0043 ***	−0.0044 ***	−0.0044 ***	−0.0438 ***	−0.0437 ***	−0.0433 ***	−0.0044 ***	−0.0016 ***
	(−16.31)	(−16.55)	(−16.46)	(−21.09)	(−21.07)	(−20.90)	(−16.50)	(−3.13)
Constant	1.2289 ***	1.2397 ***	1.2500 ***	4.3587 ***	4.3758 ***	4.5025 ***	1.1756 ***	0.3481 ***
	(12.86)	(12.97)	(13.14)	(4.89)	(4.91)	(5.10)	(12.10)	(8.55)
Observations	38,693	38,693	38,693	38,693	38,693	38,693	38,693	38,693
Adjusted R-squared	0.056	0.056	0.056	0.208	0.207	0.207	0.056	0.113
Company FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Area FE							Yes	Yes
Industry FE	Yes	Yes	Yes	Yes	Yes	Yes		Yes

Note: ***, **, and * represent significance at the 1%, 5%, and 10% levels, respectively. T statistics in parentheses.

. These findings suggest that except for the insignificance between physical risks and the credit chain, the coefficients of the other three core explanatory variables are remarkably negative at the 5% statistical level, suggesting that the conclusions of this study are generally robust.

5.2. Examining an Alternative Dependent Variable

We applied the natural logarithm of the shadow banking scale as a substitute dependent variable in our analysis to mitigate the impact of data skewness. The findings presented in columns 4–6 of Table 6 reveal a significantly negative coefficient for climate risk, substantiating that climate risks notably diminish the shadow banking activities of non-financial firms. Our conclusion remains robust with this alternative measure.

5.3. Examining an Alteration of Estimation Methods

Different industry sizes, economic policies in various years, and resource endowments in different regions can all affect a firm’s shadow banking activities and climate risks. We use different fixed effects to control unobservable heterogeneity, thus avoiding bias due to omitted variables. Table 6 shows the regression outcomes of the baseline model under different estimation methods. In column 7, after controlling for company and area fixed effects, the confidence interval coefficients are remarkably negative at the 1% level. Column 8 shows the regression results of the climate risk index and its impact on the shadow banking activities, incorporating area and year-fixed effects, with the confidence interval coefficients again remarkably negative at the 1% level. The results in Table 6 indicate a negative correlation between climate risks and shadow banking activities, supporting the predicted outcome of Hypothesis 1.

6. Further Examination

6.1. Heterogeneity Analysis Based on Financing Constraints

To delve deeper into how climate risks influence the shadow banking activities of non-financial firms, we carried out a heterogeneity analysis focusing on the aspect of financing constraints. We employed the SA index [67] as a tool to evaluate the degree of financing constraints encountered by these firms.

The construction of the SA index relies on two variables that are relatively stable over time and have strong exogeneity, namely, firm size and firm age. The formula for calculating the SA index is as follows:

$$SA=-0.737\times size+0.043\times siz{e}^2-0.04\times age$$

(2)

Here, size refers to the natural logarithm of the firm’s total assets, while age represents the firm’s age. The SA index generally takes on a negative value, with a larger absolute value indicating a higher degree of financing constraints for the firm.

The findings presented in

Table 7. Heterogeneity test: financing constraints.

	(1)	(2)	(3)	(4)	(5)	(6)
	Weak Financing Constraints			Strong Financing Constraints
Variables	SB	SB_CC	SB_CI	SB	SB_CC	SB_CI
CC1	−0.0116	−0.0003	−0.0110	−0.0257 ***	−0.0011 ***	−0.0243 ***
	(−1.56)	(−0.50)	(−1.58)	(−4.14)	(−2.86)	(−4.11)
Leverage	−0.1504 ***	−0.0112 ***	−0.1384 ***	−0.1144 ***	−0.0090 ***	−0.1049 ***
	(−7.03)	(−7.18)	(−6.83)	(−6.24)	(−7.09)	(−6.08)
Roa	−0.0684 **	−0.0003	−0.0681 **	−0.0241	−0.0012	−0.0248
	(−2.40)	(−0.14)	(−2.52)	(−0.70)	(−0.50)	(−0.76)
Size	−0.0308 ***	−0.0019 ***	−0.0288 ***	−0.0337 ***	−0.0017 ***	−0.0320 ***
	(−5.61)	(−4.12)	(−5.53)	(−7.66)	(−6.40)	(−7.59)
Lnage	−0.0326 ***	−0.0044 ***	−0.0277 ***	−0.0047	−0.0028 ***	−0.0019
	(−3.86)	(−6.97)	(−3.44)	(−0.98)	(−8.14)	(−0.40)
Dual	−0.0013	−0.0006	−0.0001	0.0242 ***	0.0010 **	0.0226 ***
	(−0.18)	(−1.25)	(−0.02)	(3.44)	(2.14)	(3.37)
Tang	−0.1328 ***	−0.0070 ***	−0.1259 ***	−0.1248 ***	−0.0067 ***	−0.1161 ***
	(−6.58)	(−4.57)	(−6.60)	(−6.32)	(−4.98)	(−6.16)
Interest	−0.0000	−0.0000	−0.0000	0.0000	0.0000	0.0000
	(−0.20)	(−0.22)	(−0.27)	(0.24)	(0.11)	(0.22)
M2	−0.3454 ***	−0.0357 ***	−0.3184 ***	−0.5955 ***	−0.0509 ***	−0.5402 ***
	(−3.84)	(−5.57)	(−3.71)	(−7.19)	(−9.33)	(−6.79)
MacroPI	0.0002	0.0000	0.0003	0.0016 ***	0.0001 ***	0.0015 ***
	(1.49)	(0.83)	(1.61)	(5.06)	(4.14)	(4.96)
BankPI	−0.0040 ***	−0.0003 ***	−0.0037 ***	−0.0047 ***	−0.0003 ***	−0.0044 ***
	(−10.96)	(−11.67)	(−10.49)	(−11.03)	(−13.06)	(−10.62)
Constant	1.3019 ***	0.0868 ***	1.2023 ***	1.2834 ***	0.0801 ***	1.2003 ***
	(9.59)	(7.96)	(9.34)	(9.39)	(10.39)	(9.14)
Observations	19,347	19,347	19,347	19,346	19,346	19,346
Adjusted R-squared	0.051	0.060	0.047	0.059	0.073	0.055
Company FE	Yes	Yes	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes	Yes	Yes

Note: T statistics in parentheses. *** and ** indicate significance at the 1%, 5% levels, respectively.

suggest that the impact of climate risks on the shadow banking activities of non-financial firms varies across groups characterized by different levels of financing constraints. In the group of firms with weaker financing constraints, there is no significant correlation between climate risks and shadow banking activities. However, in the group with stronger financing constraints, there is a significant negative correlation. This suggests that when climate risks increase, commercial banks reduce credit lending, leading to a greater reduction in shadow banking activities for firms that are more heavily constrained in their financing.

6.2. Heterogeneity Analysis Based on Profitability

We further categorize firms into two groups based on profitability, using the median value of Roa to distinguish between firms with weak and strong profitability. This allows us to delve deeper into how climate risks affect the shadow banking activities of firms with different levels of profitability.

Outcomes in

Table 8. Heterogeneity test: profitability.

	(1)	(2)	(3)	(4)	(5)	(6)
	Weak Profitability			Strong Profitability
Variables	SB	SB_CC	SB_CI	SB	SB_CC	SB_CI
CC1	−0.0277 ***	−0.0009	−0.0258 ***	−0.0103 *	−0.0004	−0.0097 *
	(−3.26)	(−1.44)	(−3.19)	(−1.65)	(−0.95)	(−1.66)
Leverage	−0.1298 ***	−0.0095 ***	−0.1199 ***	−0.1117 ***	−0.0104 ***	−0.1010 ***
	(−5.86)	(−6.14)	(−5.64)	(−5.93)	(−7.24)	(−5.71)
Roa	−0.2225 ***	−0.0157 ***	−0.2094 ***	−0.0728 ***	−0.0002	−0.0731 ***
	(−3.32)	(−3.42)	(−3.26)	(−2.74)	(−0.11)	(−2.89)
Size	−0.0517 ***	−0.0023 ***	−0.0492 ***	−0.0209 ***	−0.0013 ***	−0.0194 ***
	(−8.51)	(−6.14)	(−8.47)	(−4.87)	(−3.81)	(−4.75)
Lnage	−0.0004	−0.0034 ***	0.0033	−0.0302 ***	−0.0029 ***	−0.0270 ***
	(−0.07)	(−8.44)	(0.60)	(−4.86)	(−6.90)	(−4.55)
Dual	0.0191 **	0.0004	0.0180 **	0.0151 ***	0.0002	0.0149 ***
	(2.47)	(0.83)	(2.42)	(2.65)	(0.56)	(2.75)
Tang	−0.1863 ***	−0.0086 ***	−0.1758 ***	−0.1221 ***	−0.0062 ***	−0.1164 ***
	(−7.05)	(−5.10)	(−6.97)	(−6.81)	(−4.75)	(−6.84)
Interest	0.0000	−0.0000	−0.0000	−0.0000	0.0000	−0.0000
	(0.02)	(−0.11)	(−0.02)	(−0.38)	(0.88)	(−0.49)
M2	−0.6887 ***	−0.0555 ***	−0.6325 ***	−0.3662 ***	−0.0317 ***	−0.3346 ***
	(−7.89)	(−9.88)	(−7.53)	(−5.87)	(−6.59)	(−5.64)
MacroPI	0.0011 ***	0.0001 ***	0.0011 ***	0.0005 ***	0.0000 *	0.0005 ***
	(4.31)	(3.48)	(4.25)	(3.11)	(1.68)	(3.07)
BankPI	−0.0059 ***	−0.0004 ***	−0.0054 ***	−0.0031 ***	−0.0002 ***	−0.0028 ***
	(−12.62)	(−15.12)	(−12.03)	(−10.48)	(−10.79)	(−10.09)
Constant	1.8136 ***	0.0885 ***	1.7131 ***	0.9793 ***	0.0699 ***	0.9027 ***
	(11.73)	(6.55)	(11.51)	(8.14)	(7.38)	(7.90)
Observations	19,346	19,346	19,346	19,347	19,347	19,347
Adjusted R-squared	0.063	0.077	0.058	0.051	0.054	0.048
Company FE	Yes	Yes	Yes	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes	Yes	Yes	Yes

Note: ***, **, and * represent significance at the 1%, 5%, and 10% levels, respectively. T statistics in parentheses.

indicate differences in the relationship between climate risks and shadow banking activities across groups with different levels of profitability. Specifically, for firms with weak profitability (columns 1–3), the coefficient of climate risk is negative and remarkable. Contrarily, in firms with strong profitability (columns 4–6), the coefficient of climate risk is not remarkable. This suggests that the impact of climate risks on shadow banking activities is more pronounced in firms with poorer profitability. This finding implies that non-financial firms with weaker profitability might be more vulnerable to the effects of climate risks, thereby reducing their shadow banking activities. Notably, the effect of climate risks on the credit intermediary activities of companies with weak profitability is more significant. This indicates that firms with weaker profitability are more likely to be affected by climate change, leading to deteriorated credit conditions and, consequently, increased costs and risks associated with their role as credit intermediaries.

7. Mechanism Analysis

The preceding analyses offer direct evidence that rising climate risks can suppress the shadow banking activities of non-financial firms. We delve into the underlying mechanisms behind this relationship in this part.

Considering that an increasing number of studies indicate that the traditional three-stage mediation effect might have significant flaws [68,69], specifically, the three models involving three sets of variables could potentially face three endogeneity issues, which would require a minimum of two instrumental variables (IV1 for X→Y and X→M; IV2 for M→Y) and the assumption that the three error terms (e1, e2, and e3) are mutually uncorrelated. Since empirical research often relies on observational data, addressing these endogeneity issues can make the analysis highly complex. To resolve this, our study primarily adheres to the design method of Aguinis [68]. First, we increase the use of bootstrap-derived percentile confidence intervals, which is able to broaden the hypotheses in the Sobel test, where the mediation effect rests on the premise that the multiplication of coefficients follows a normal distribution [70]. Second, we consider the link between the mediator and dependent variables in the mediation effect, aiding in enhancing the integrality of the empirical chain. Based on this, our study draws on the approach of Niu and employs a four-stage mediation effect model for testing [71]. Following the four-stage mediation effect approach analyzed previously and in conjunction with the design of model (1) from earlier, we construct the following mediation effect models (2), (3), and (4):

$$M={\beta }_0+{\beta }_1{CC1}_{i,t}+{\beta }_2{Z}_{i,t}+{\delta }_t+{\phi }_i+{\epsilon }_{i,t}$$

(3)

$${SB}_{i,t}={\beta }_0+{\beta }_1{M}_{i,t}+{\beta }_2{Z}_{i,t}+{\delta }_t+{\phi }_i+{\epsilon }_{i,t}$$

(4)

$${SB}_{i,t}={\beta }_0+{\beta }_1{CC1}_{i,t}+{\beta }_2{M}_{i,t}+{\beta }_2{Z}_{i,t}+{\delta }_t+{\phi }_i+{\epsilon }_{i,t}$$

(5)

In the aforementioned models, $M$ represents the mediator variable, which in this case are the cross-risks and cash holding preferences. The steps for testing the mediation effect are as follows: First, test whether ${\beta }_1$ in Equation (1) is significant; if it is significant, continue with the mediation effect test. Next, test whether ${\beta }_1$ in Equation (3) is significant, and also test whether ${\beta }_1$ in Equation (4) is significant. If both are significant, further test ${\beta }_1$ in Equation (5). If this is significant, it indicates a partial mediation effect; if it is not significant, it indicates a full mediation effect. If ${\beta }_1$ in either Equation (3) or Equation (4) is not significant, a Sobel test is required. A significant Z value in the Sobel test indicates a significant mediation effect; otherwise, the mediation effect is not significant.

7.1. Expansion of Cross-Risk

We examine whether the amplification of cross-risks serves as a potential mediating mechanism through which climate risks influence the shadow banking operations of non-financial companies. Diversified businesses typically involve multiple operational areas and have complex business connections and balance sheet entanglements with other enterprises, forming a tightly interconnected financial network. When such diversified firms face climate risks, these risks can be transmitted to other businesses through this complex network of business interconnections, thereby amplifying and impacting the entire system. This can lead to firms engaged in high-risk, high-leverage, and informationally asymmetrical shadow banking being unable to recover funds on time, increasing the risks of liquidity crises and stock price crashes. In summary, under the backdrop of climate risks, firms with a higher degree of business diversification, due to the interconnectedness of their operations and balance sheets, are more likely to trigger a more substantial cross-contagion effect with other businesses. Therefore, these firms tend to reduce their shadow banking activities.

Following the approach used by Han [65], we measure a firm’s cross-risks based on whether the number of business areas covered by its main operating revenue exceeds or falls below the industry’s yearly average. When a firm’s main operating revenue covers more business areas than the average in its industry, we mark it as 1, indicating the presence of cross-risks; otherwise, it is marked as 0.

The detailed outcomes of the regression are presented in

Table 9. Channels through which climate risks effect shadow banking activities: cross-risks.

	(1)	(2)	(3)
Variables	RISK2	SB	SB
CC1	0.0621 ***		−0.0223 ***
	(4.20)		(−4.65)
RISK2		−0.0033	−0.0029
		(−0.96)	(−0.84)
Leverage	−0.0116	−0.1193 ***	−0.1197 ***
	(−0.35)	(−8.17)	(−8.20)
Roa	−0.1285 ***	−0.0543 **	−0.0517 **
	(−2.68)	(−2.36)	(−2.24)
Size	0.0506 ***	−0.0315 ***	−0.0306 ***
	(6.00)	(−8.82)	(−8.53)
Lnage	0.0671 ***	−0.0143 ***	−0.0134 ***
	(8.40)	(−3.64)	(−3.38)
Dual	0.0178 *	0.0168 ***	0.0166 ***
	(1.89)	(3.37)	(3.33)
Tang	0.0126	−0.1324 ***	−0.1325 ***
	(0.38)	(−9.22)	(−9.23)
Interest	−0.0001 **	0.0000	0.0000
	(−2.04)	(0.27)	(0.27)
M2	0.2166 **	−0.5111 ***	−0.5027 ***
	(2.09)	(−9.74)	(−9.57)
MacroPI	0.0019 ***	0.0009 ***	0.0008 ***
	(6.81)	(6.36)	(5.98)
BankPI	0.0020 ***	−0.0044 ***	−0.0044 ***
	(4.66)	(−16.46)	(−16.56)
Constant	−1.2101 ***	1.2489 ***	1.2353 ***
	(−6.07)	(13.20)	(13.00)
Sobel Z	−2.12 **
Bootstrap (1000 replicates) Testing confidence intervals	[−0.0007, −0.00003]
Observations	38,693	38,693	38,693
Adjusted R-squared	0.025	0.056	0.056
Company FE	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes

Note: ***, **, and * represent significance at the 1%, 5%, and 10% levels, respectively. T statistics in parentheses.

. In column 1, the coefficient of CC1 is remarkably positive, suggesting a remarkable positive link between climate risks and cross-risks. In column 3, the coefficient of CC1 is remarkably negative, indicating a remarkable negative link between climate risks and shadow banking activities, with the absolute value of the CC1 coefficient decreasing compared to the benchmark regression when using a stepwise regression method. Based on these findings, further Sobel tests were conducted, revealing a Z statistic of −2.12, significant at the 5% level. Additionally, this study conducted a Bootstrap (1000 replicates) test, finding that the 95% confidence interval for the mediation effect is [−0.0007, −0.00003], which does not include 0. These results indicate that cross-risks have a mediating effect. That is, climate risks diminish the shadow banking operations of non-financial companies by increasing corporate cross-risks, thus validating Hypothesis 1 of this study.

7.2. Preference for Cash Holding

In this study, we further examine the mediating mechanism of the preference for cash holding. First, climate change-related extreme weather and abnormal temperatures not only deteriorate a firm’s tangible assets but also increase operational costs [72,73], thereby elevating cash flow risks and increasing the risk of financial default. Existing research shows that companies choose to keep more cash on hand to mitigate external risk impacts [74,75,76], and thus, an increase in climate risks can trigger a firm’s preference for holding cash. The definition of a firm’s cash holding is as follows: measurement of annual changes in cash holdings, scaled by total assets.

The regression data are presented in

Table 10. Channels through which climate risks effect shadow banking activities: cash holding.

	(1)	(2)	(3)
Variables	CASH	SB	SB
CC1	0.0152 ***		−0.0193 ***
	(4.08)		(−4.02)
CASH		−0.2122 ***	−0.2109 ***
		(−10.36)	(−10.31)
Leverage	−0.2038 ***	−0.1625 ***	−0.1626 ***
	(−18.19)	(−10.61)	(−10.62)
Roa	0.0676 ***	−0.0392 *	−0.0371
	(4.45)	(−1.71)	(−1.61)
Size	0.0114 ***	−0.0292 ***	−0.0283 ***
	(4.25)	(−8.08)	(−7.84)
Lnage	−0.0388 ***	−0.0226 ***	−0.0218 ***
	(−15.27)	(−5.59)	(−5.33)
Dual	−0.0013	0.0164 ***	0.0162 ***
	(−0.51)	(3.30)	(3.26)
Tang	−0.3681 ***	−0.2106 ***	−0.2101 ***
	(−32.09)	(−12.38)	(−12.37)
Interest	−0.0001 ***	−0.0000	−0.0000
	(−3.55)	(−0.08)	(−0.09)
M2	0.3372 ***	−0.4391 ***	−0.4322 ***
	(12.21)	(−8.16)	(−8.03)
MacroPI	−0.0001 *	0.0008 ***	0.0008 ***
	(−1.89)	(6.09)	(5.76)
BankPI	0.0017 ***	−0.0040 ***	−0.0040 ***
	(14.62)	(−15.93)	(−16.03)
Constant	0.0413	1.2597 ***	1.2475 ***
	(0.61)	(13.19)	(13.01)
Sobel Z	−6.74 ***
Bootstrap (1000 replicates) Testing confidence intervals	[−0.0021, −0.0010]
Observations	38,693	38,693	38,693
Adjusted R-squared	0.291	0.067	0.068
Company FE	Yes	Yes	Yes
Industry FE	Yes	Yes	Yes

Note: T statistics in parentheses. ***, * indicate significance at the 1%, 10% levels, respectively.

. In column 1, the coefficient of CC1 is remarkably positive, indicating a remarkable positive link between climate risks and cash holdings. Column 2 shows a significant positive correlation between cash holdings and the shadow banking activities of non-financial firms. In column 3, the coefficient of CC1 is remarkably negative, demonstrating a significant negative relationship between climate risks and shadow banking activities. Additionally, the absolute value of the CC1 coefficient has increased compared to the benchmark regression when using the stepwise regression method. Based on these findings, a further Sobel test was conducted, yielding a Z statistic of −6.74, which is remarkable at the 1% level. A Bootstrap test with 1000 replicates also indicated that the 95% confidence interval for the mediation effect is [−0.0021, −0.0010], which does not include 0. These outcomes show that cash holdings possess a mediating effect, confirming that climate risks decrease the shadow banking operations of non-financial firms by increasing the cash reserves, thereby supporting Hypothesis 1 of this study.

8. Conclusions and Policy Recommendations

Our empirical results address the two questions posed earlier. First, the findings demonstrate that climate risks possess a markedly adverse effect on the shadow banking operations of non-financial firms. Specifically, as climate risks increase, there is a significant reduction in the extent of shadow banking activities conducted by these companies.

Second, this study identifies the mechanisms through which climate risks influence shadow banking activities. It shows that climate risks indirectly reduce non-financial companies’ engagement in shadow banking by amplifying cross-risk exposure and increasing their cash holdings. Climate risks enhance cross-risk exposure by introducing additional sources of risk, which affects firms’ risk tolerance and financial strategies. To mitigate the uncertainties and potential financial losses associated with climate change, firms are likely to reduce their involvement in high-risk shadow banking activities. Additionally, to address potential risks relating to climate, companies tend to have larger cash reserves, directly decreasing their reliance on shadow banking. Together, these elements lead to the observed decrease in shadow banking operations among non-financial firms.

Unlike previous studies that pay close attention to the activities of commercial banks not listed on the balance sheet, this research examines the “quasi-financial” behaviors of non-financial firms, exploring how these firms adjust their shadow banking activities as a response to climate risks. These findings reveal that rising climate risks lead to a reduction in the shadow banking operations of non-financial companies. On one hand, climate risks exacerbate cross-risks and expand sources of risk for companies, while on the other hand, to manage potential threats, firms increase their preference for holding cash. These two factors together contribute to the decline in shadow banking activities. The heterogeneity analysis shows that firms with weaker profitability and greater financing constraints are more sensitive to climate risks.

Drawing from findings, the research suggests these policy recommendations: First, both governments and enterprises need to be vigilant about the potential risk expansion brought about by climate risks. The research results suggest that the reason for non-financial firms reducing their shadow banking activities lies in the heightened uncertainty about the future economic environment due to climate risks, forcing them to embrace a more prudent strategy in investing as well as financial strategies involving high-risk and opaque shadow banking products. Therefore, for governments, formulating policies to respond to climate risks should focus not only on visible financial risks (such as the increased operating costs due to physical risks) but also on avoiding the creation of new risks in the process of risk prevention. Especially under economic downturn pressures, while reasonably boosting corporate confidence, it is important to prevent issues like the aggregation and interconnection of risks from a policy operation perspective. For enterprises confronted with climate risk challenges, they should develop more proactive and effective strategic transformations and supporting plans, allocating limited capital to their main businesses that have long-term developmental significance. This approach is critical to avoid pursuing high-risk shadow banking activities with relatively higher returns and greater obscurity due to the lower short-term returns of the real economy, especially to avoid falling into adverse cycles caused by intertwined risks.

Second, government policymakers should consider the additional impacts brought by climate risks when deepening financial reforms. For governments, it is crucial to improve the investment environment for real businesses, actively guiding the economic focus of non-financial firms from a “virtual economy” to a “real economy”. This means reversing the trend of capital shifting from tangible to intangible assets, enhancing investments in firms’ main operations and profits from these operations, which is important for reducing risk-bearing in extreme situations. When economic pressures caused by climate risks are significant, regulatory authorities should consider moderately increasing tolerance towards shadow banking operations of non-financial companies. This would encourage them to proactively disclose risk information, enabling regulatory bodies to take further measures to mitigate potential financial risks and avoid systemic financial risks and crises. At the same time, it would enhance the resilience of enterprises to climate risks.

Lastly, for enterprises, it is imperative to proactively reduce high-risk underground financial investment and financing activities and to improve the mechanism for disclosing climate risk information. The proverb “It is better to solve the root problem rather than treating the symptoms” aptly applies here. Allocating limited capital to main businesses with long-term developmental significance and enhancing operational resilience and capabilities are vital to better manage risk mutations in extreme situations. Our study’s findings indicate that enterprises with weaker profitability and stronger financing constraints are more sensitive in responding to climate risks. These more sensitive and vulnerable firms should especially avoid engaging in high-risk shadow banking operations. What is more, companies should enhance their mechanisms for disclosing climate risk information, thereby increasing corporate sensitivity to climate risks and curbing the possibility of risk accumulation among various stakeholders.

Despite providing a preliminary theoretical discussion and analysis of the relationship between climate risks and shadow banking activities among non-financial companies, the research has limitations that require further improvement. First, the methodology primarily relies on textual analysis for statistical evaluation and does not utilize more objective data such as carbon footprint or carbon emissions-adjusted loan sizes. This limitation stems mainly from the availability and accessibility of detailed data. We acknowledge this as a constraint of our study and hope to delve deeper into such analyses when data acquisition conditions improve. Second, our research is predominantly grounded in the data from Chinese listed companies, which may limit the general applicability of our conclusions. While these companies hold a significant position in the financial market, they may differ considerably from non-listed companies or businesses in other countries in terms of operations, financing structures, and market behavior. Thus, our findings may not fully apply to a broader group of non-financial firms or economic environments in other countries. Future research should consider employing a more diverse sample to enhance the universality and depth of the findings.