Climate risk and corporate green transformation: nonlinear evidence from Chinese listed firms

Against the backdrop of intensifying climate-related risks and environmental pressures, corporate green transformation has become an increasingly important response for improving environmental performance and promoting sustainable development (Ambec et al., 2013). Corporate green transformation refers to the process by which enterprises reshape their strategies and operations around the constraints of the natural environment, manifested in a profound transformation from pollution prevention and process improvement to product responsibility and sustainable development (Aragón-Correa and Sharma, 2003). With the continuous improvement of policy guidance and market mechanisms, developed economies have made significant progress in the decarbonization of industrial structures and the diffusion of clean technologies (Aghion et al., 2016). Green production investments by European Union firms are reflected not only in reduced pollution intensity but also in adjustments to technological pathways (Colmer et al., 2025). As the world’s largest carbon emitter, China’s green transformation is at a critical stage of shifting from end-of-pipe treatment to a systemic reshaping of the entire industrial chain (Lu et al., 2023). Under the constraints of dual carbon targets, corporate green transformation has become a crucial support for China’s sustainable development (Yu et al., 2021). Environmental regulatory policies (Gao and He, 2021), digitalization levels (Feng et al., 2024), green credit (Zhang et al., 2024), digital finance and the flow of innovative factors can all significantly influence corporate green transformation decisions. However, because green transformation typically involves high upfront investment and long payback periods, and its outcomes are highly uncertain, companies often face a realistic trade-off between profit maximization and long-term green investment. Whether corporate green strategies will continue to advance when external environmental risks rise remains to be further examined (Wang and Cao, 2022).

In recent years, climate risk has gradually become an important variable explaining the differences in corporate green transformation. Climate risk refers to the possibility that the uncertainty caused by climate change will negatively impact natural systems, human societies and economic systems (Klein et al., 2014). The World Economic Forum’s Global Risks Report 2025 identifies climate risk as a key factor constraining economic transformation and corporate investment decisions (World Economic Forum, 2025). Climate risk typically includes transition risk and physical risk. Transition risk stems from rapid changes in policies, regulations and market preferences during the transition to a low-carbon economy, while physical risk arises from the direct impact of extreme weather events and long-term climate change on corporate assets and operations (Financial Stability Board, 2020). Scholars have examined the impact of climate risk on capital pricing, corporate innovation activities (Li et al., 2024a) and operating performance. Regarding environmental impact, Ren et al. (2022) found that climate risk increases corporate carbon emissions, with the increase being more pronounced when climate risk is in a high-risk range. Wang et al. (2024a) found that corporate disclosure of climate risk information can reduce carbon emission levels. Climate risk may incentivize firms to accelerate their green transformation by strengthening environmental constraints and increasing compliance pressure (Porter and Linde, 1995). However, excessive uncertainty may also suppress firms’ long-term investment intentions (Bloom et al., 2007), leading to the postponement or reduction of green investments. Existing research mainly examines the impact of climate risk on corporate green transformation from a linear framework, with limited exploration of its potential nonlinear characteristics. Therefore, this paper examines the nonlinear impact of climate risk on corporate green transformation, providing new empirical evidence for understanding the decision-making logic of corporate green transformation in the context of climate risk.

The contributions of this paper are as follows. First, this paper extends the literature on the economic consequences of climate risk by identifying an inverted U-shaped association between climate risk and corporate green transformation. Previous studies mainly estimate their average linear relationship (Fang, 2024; Jin and Gao, 2025), whereas this study further shows that the dominant mechanisms change across risk levels. Second, this paper incorporates the competition mechanism and capital reallocation logic into a unified analytical framework, deepening the theoretical explanation of the impact of climate risk on corporate green transformation from the dimensions of impact transmission mechanisms and corporate adaptability. Third, this paper expands the research on the impact of climate risk on corporate behavior from the perspective of differences in risk types and corporate strategic responses. Under high climate risk, companies’ symbolic green disclosure continues to increase, while substantive green transformation weakens, exhibiting a clear divergence between symbolic disclosure and substantive green transformation.

The remainder of this paper is organized as follows. Section 2 develops the theoretical framework and proposes four research hypotheses. Section 3 describes the models, variables and data. Section 4 shows the empirical results. Section 5 discusses the theoretical and practical implications while addressing the research limitations. Finally, Section 6 summarizes the main findings.

The Porter hypothesis emphasizes that moderate environmental constraints can enhance a firm’s competitive advantage through innovation compensation mechanisms (Porter and Linde, 1995). When external environmental constraints intensify, firms will improve resource efficiency through technological innovation and production method upgrades to offset compliance costs to meet regulatory expectations and market demands (Ambec et al., 2013). Environmental constraints may drive a firm’s sustainable development capabilities, thereby benefiting it in long-term competition. Environmental regulations can also promote green technology innovation and patent output (Lanoie et al., 2011), and improve firm productivity (Albrizio et al., 2017). As climate change has become a core issue in global governance, capital markets have shown significantly increased attention to climate risk disclosure (Krueger et al., 2020), environmental, social and governance investment has continued to expand (Friede et al., 2015), and investors’ demands for risk premiums for high-carbon companies have been rising (Bolton and Kacperczyk, 2021). Climate risk, through capital market pricing and regulatory expectations, forms a mechanism that reinforces institutional signals and market expectations, forcing companies to respond proactively through investment and green innovation (Li et al., 2024b). When companies interpret climate risk as increased future environmental regulations, their subjective expectations regarding future compliance costs and changes in market structure will adjust accordingly, reflected in their financial and operational decisions (Dang et al., 2023). Furthermore, the implementation of policies such as carbon emissions trading will significantly alter corporate costs and resource allocation, reinforcing companies’ strategic trade-offs regarding low-carbon strategies (Bartram et al., 2022). Companies that proactively invest in green technologies gain a first-mover advantage (Lieberman and Montgomery, 1988). Therefore, within the low to moderate climate risk range, the anticipated institutional strengthening and shifting market preferences resulting from increased climate risk may enhance the expected returns on corporate green investments. At this stage, climate risk primarily acts as a stimulus effect, thereby reinforcing companies’ motivation to engage in green innovation and structural upgrading.

Real options theory posits that when investments are irreversible and future returns are highly uncertain, companies face significant waiting value (Bernanke, 1983). Green transformations typically involve highly specific capital investments and long-term research and development expenditures, which are difficult to fully withdraw once implemented (Pindyck, 1990). When climate risk rises significantly, the uncertainty of future corporate returns increases, and immediately investing in highly specific, long-term green transformations means forgoing the option value of waiting for clearer signals (Sun et al., 2024). The inhibitory effect of uncertainty shocks on corporate investment has been widely verified at both macro and micro levels. Bloom (2009) found that increased uncertainty significantly delays corporate capital expenditure decisions. Gulen and Ion (2016) found that policy uncertainty significantly reduces corporate capital expenditure levels. In the scenario of rising climate risk, risk management motives may prompt firms to increase cash holdings and liquidity reserves, thereby creating a crowding-out effect on long-term green investment (Li et al., 2024a).

Porter’s hypothesis emphasizes the innovation incentive effect brought about by environmental constraints, while real options theory emphasizes the investment inhibition effect brought about by increased uncertainty. These two mechanisms may shift in dominance across different ranges as climate risk levels change, as shown in Figure 1. The incentive effect gained by firms from green transformation may gradually increase with rising climate risk, while the crowding-out effect caused by uncertainty, investment irreversibility and resource constraints may increase at a faster rate. The difference between the two, that is, the net impact of corporate green transformation, may exhibit an inverted U-shaped relationship of first increasing and then decreasing. Thus, this paper proposes H1:

Based on management attention theory, rising external environmental pressures may prompt top management teams (TMTs) to reallocate their limited attention resources (Ocasio, 1997). When climate risk is at a low-to-medium level, gradually increasing climate pressure is often identified by TMTs as a forward-looking risk signal, prompting enterprises to focus on potential supply chain vulnerabilities and disruption risks in their business decisions and to diversify risk exposure by adjusting supply chain configurations. Increased TMT attention to environmental issues also helps to promote discussions on green transformation within enterprises, strengthen the shared understanding of environmental issues within the organization and enhance the motivation for green innovation (Huang et al., 2024). In this stage, firms tend to transform external pressure into internal transformation drivers, enhancing organizational resilience and coping capabilities through green technology innovation and process upgrades (Wang et al., 2024b). However, according to real options theory, when climate risks enter a high-uncertainty or even extreme range, firms tend to postpone investment decisions (Wreford et al., 2020), reflecting a greater inclination toward cautious capital allocation strategies in highly uncertain environments (Zhao et al., 2025). While supply chain uncertainty persists, firms may become more cautious in allocating resources to green transformation. Therefore, at low levels of climate risk, heightened managerial perception of supply chain uncertainty is expected to translate external climate signals into stronger incentives for green transformation, whereas this mechanism weakens as climate risk enters the high-risk range. Accordingly, H2 is proposed:

When facing climate risk exposure, firms exhibit observable differences in their responses in investment, green innovation and organizational adjustments, providing a micro-foundation for the evolution of industrial competitive structures caused by climate risk (Li et al., 2024b). In a more competitive market environment, external regulatory pressure and product market competition are complementary, significantly incentivizing companies to engage in green innovation to maintain or enhance their competitive position (Dai et al., 2025). However, climate vulnerability and related ecological risks significantly increase the risk of financial distress and bankruptcy for companies. This means that as pressure gradually increases, the survival constraints of more vulnerable companies within the industry tighten, and resources and market share are more likely to be reallocated to relatively robust and resilient companies (Li et al., 2024a). When climate risks continue to rise and enter a period of high uncertainty or even extreme conditions, companies face significantly greater uncertainty regarding irreversible investments and policy paths, leading to more cautious entry and expansion (Wreford et al., 2020). Climate policy uncertainty (CPU) triggers structural mergers and acquisitions or resource restructuring (Yang et al., 2025), resulting in market share concentration toward more productive companies. Therefore, at low levels of climate risk, intensified market competition is expected to convert climate pressure into stronger incentives for green innovation and transformation. However, as climate risk rises further, uncertainty and survival pressure weaken this competitive channel, making it less effective in promoting green transformation. Accordingly, H3 is proposed:

When climate risk is low, climate exposure is more often seen as a long-term, gradual risk rather than a shock factor that immediately affects asset pricing (Bolton and Kacperczyk, 2021). Long-term institutional investors also do not need to significantly adjust their portfolios. In this stage, the impact of climate risk on patient capital may not be significant. When climate risk enters the high-uncertainty range, climate risk begins to affect the capital supply structure through two channels: asset pricing and portfolio reallocation. Companies with high climate risk exposure face higher risk premiums and tail risk pricing (Ilhan et al., 2021). Corporate stock prices become significantly more sensitive to climate risk factors (Bolton and Kacperczyk, 2021), and the market tends to be pessimistic. After risk exposure becomes apparent, institutional investors often adjust their investment structure and reduce their allocation to companies with high climate risk to control the overall risk exposure of their portfolios (Blanco et al., 2024). Since patient capital provides stable financial support for long-term innovation and green transformation (Flammer, 2021), its contraction weakens firms’ capacity to sustain irreversible green investment. As long-term financing declines, firms become more inclined to preserve liquidity and postpone investment, thereby increasing the value of waiting. Therefore, in the high-risk stage, climate risk is expected to inhibit green transformation through the contraction of patient capital and stronger waiting-option incentives. Accordingly, H4 is proposed:

This study constructs the following two-way fixed-effects (FE) model to examine the nonlinear impact of climate risk on corporate green transformation (Wooldridge, 2010):

$GTransi,t$ represents the level of green transformation of firm i in year t. $ClimateRiski,t$ denotes the firm’s climate risk exposure. ClimateRisk and its squared term jointly capture the nonlinear effect of climate risk on corporate green transformation. $Zi,t$ is the control variable. $μi$ and $ηt$ represent the firm FE and the year FE, respectively, used to control for individual characteristics that do not change over time and the impact of macro-environmental fluctuations over time. $εi,t$ represents the random disturbance term. Standard errors are clustered at the firm level to account for within-firm serial correlation and heteroskedasticity.

The quadratic specification is adopted for both theoretical and empirical reasons. Theoretically, Porter’s Hypothesis predicts an initially positive incentive effect, whereas Real Options Theory predicts an increasingly negative investment-deterring effect as uncertainty rises. Their combination naturally implies a smooth turning point rather than a discrete threshold. Empirically, the quadratic model provides a parsimonious representation of this pattern and allows the turning point and marginal effects to be directly estimated. To assess whether the results are driven by this functional-form assumption, we further include a cubic term and conduct U-shape tests in the robustness analysis.

The dependent variable in this paper is corporate green transformation (GTrans). To characterize the substantive actions of corporate green transformation, this paper uses the entropy method to comprehensively weight and measure various indicators. This paper constructs an evaluation system from the dimensions of green technology innovation, green production efficiency, environmental governance actions and environmental end-use performance (Fang, 2024; Niu and Wang, 2024; Zhou et al., 2024). The number of granted green patents and their citation counts are used to measure corporate green technology innovation. Whether or not the company has passed ISO 14001 environmental management system certification and the company’s green investment expenditure reflect the company’s environmental governance actions. Environmental penalties are used to characterize the company’s environmental end-use performance. Green total factor productivity, used to measure green production efficiency, is calculated using the slacks-based measurement method (Tone, 2001). The input variables for green total factor productivity include labor input (number of employees), capital input (net fixed assets) and energy input (total energy consumption). The desirable output is operating revenue. The undesirable outputs include industrial sulfur dioxide emissions, industrial particulate matter emissions and industrial wastewater emissions.

Our independent variable is corporate climate risk exposure (ClimateRisk). Because directly translated English dictionaries may not fit the Chinese annual-report context (Li et al., 2024b), we combine terminology from prior studies and the China Meteorological Disaster Yearbook and further refine the dictionary using the corpus of Chinese A-share firms’ annual reports from 2010 to 2024, yielding a dictionary of 98 keywords covering transition and physical risks, as shown in Table 1. Compared with existing text-based measures, our indicator places greater emphasis on firm-level meteorological and hydrological exposure, such as waterlogging, freezing damage and flood conditions. ClimateRisk is calculated as the frequency of climate-risk keywords divided by the total number of words in the annual report and multiplied by 100. Its validity is further assessed using separate transition and physical-risk measures and an adjusted index excluding green-related keywords. As a text-based measure, ClimateRisk captures firms’ disclosed climate risk exposure and managerial attention rather than realized physical damage alone.

This paper selects control variables including firm size (size), leverage (lev), return on assets (roa), firm age (lnage), the shareholding ratio of the largest shareholder (top1), state ownership (soe) and fixed-asset ratio (fixasset) (Zhou et al., 2024; Jin and Gao, 2025).

This study uses Chinese A-share listed firms from 2010 to 2024. Firm-level financial and governance data are obtained from CSMAR and Wind, green patent data from CNRDS, environmental penalty data from CCER, pollution-related search data from Baidu Index, extreme weather data from the China Meteorological Administration and other macro data from city and provincial statistical yearbooks. Financial firms are excluded because of their distinct balance-sheet structures and regulatory regimes. ST, *ST, particular transfer and insolvent firms are removed because their investment and disclosure decisions may be dominated by survival and delisting pressures, while observations with missing key variables are excluded because complete data are required for estimation. All continuous variables are winsorized at the 1st and 99th percentiles. The final sample contains 43,309 firm-year observations, and Table 2 reports the descriptive statistics.

Table 3 reports the results of the benchmark regression. Columns (1) and (2) show that the coefficient of ClimateRisk is significantly positive at the 1% level, indicating that climate risk is positively associated with the green transformation of enterprises. Columns (3) and (4) introduce ClimateRisk². The results show that the coefficient of ClimateRisk is significantly positive, the coefficient of ClimateRisk² is significantly negative and the U-test passes the 1% significance test, indicating that the inverted U-shaped relationship between the variables holds. H1 is supported. As shown in Figure 2, the turning point is approximately 0.343; more than 4,000 firm-year observations, about 10% of the sample, lie above it, where the marginal association turns negative. At the sample mean of 0.154, a one-standard-deviation increase in climate risk is associated with an increase in green transformation of 20.1% of its standard deviation, indicating noticeable adjustments in firms’ green investment and operations. Beyond the turning point, heightened uncertainty, financing pressure and investment irreversibility prompt firms to preserve liquidity and postpone green investment, with the negative marginal effect reaching an absolute magnitude of 38.0% of the standard deviation of green transformation at high-risk levels. Compared with existing literature (Fang, 2024; Jin and Gao, 2025), our results are consistent with prior evidence that climate risk is positively associated with green transformation on average, but further reveal substantial nonlinearities.

To mitigate the issues of omitted variables and endogeneity caused by bidirectional causality, this paper uses the two-stage least squares (2SLS) method. Two instrumental variables (IVs) are selected. Based on the standard definition of extreme weather by Han et al. (2018), this paper selects the proportion of extreme weather events in other cities within the same province as the instrumental variable (IV1), and calculates the mean of other cities within the same province to capture regional climate impacts. This paper selects the urban air circulation coefficient as the second IV (IV2), constructed by multiplying 10-m wind speed with boundary layer height data (Jacobson, 2002). Regarding the exclusion restriction, extreme weather events in other cities within the same province are used instead of those occurring in firms’ own cities to reduce the possibility of direct impacts on firms’ production and investment decisions. Conditional on firm FEs, province-by-year FEs and firm-level controls, such regional climate shocks are expected to affect corporate green transformation mainly through their influence on perceived climate risk. The air circulation coefficient mainly captures exogenous atmospheric dispersion conditions rather than firm-level environmental decisions. These meteorological characteristics are largely determined by natural conditions beyond firms’ control and have been used as exogenous sources of variation in environmental and economic research (Hering and Poncet, 2014; Cai et al., 2016). Although the exclusion restriction cannot be directly tested, the above arguments and FEs specification provide support for the validity of the instrumental variables.

Table 4 reports the 2SLS estimates. Standard errors are triple-clustered at the firm, province and year levels. The Kleibergen–Paap rk Wald F statistic is 38.167, exceeding the Stock–Yogo critical value of 7.03, indicating that weak-instrument concerns are limited. Although the 2SLS coefficients are larger than the baseline estimates, their magnitudes should not be compared directly. The baseline model captures the average relationship in the full sample, whereas the IV estimation exploits exogenous variation in climate risk induced by extreme weather and atmospheric circulation conditions and thus identifies stronger local responses among firms that are more sensitive to these shocks (Angrist et al., 1996). In addition, measurement noise in the text-based climate risk indicator may attenuate the baseline estimates toward zero, while the IV approach partially mitigates this problem. The turning point estimated by 2SLS is 0.304, close to the baseline value of 0.343, indicating that the main findings remain robust after accounting for potential endogeneity.

To assess whether the baseline results are driven by unobserved factors or random variation, this paper conducts the placebo test (Chetty et al., 2009). While keeping the control variables constant, we perform random permutation on the independent variable. This process is repeated 1,000 times to obtain the empirical distribution of the coefficients. Figures 3 and 4 report the random simulated distributions of β1 and β2, respectively. The estimated coefficients generated by the random permutation exhibit a normal distribution centered at 0, indicating that the estimated inverted U-shaped relationship is unlikely to be driven by random assignment, thus demonstrating the robustness of the results.

This paper also conducts a series of robustness tests, the results of which are shown in Table 5. We introduce a double machine learning model, using the Lasso algorithm to mitigate model specification bias (Column 1). To eliminate the potential interference of indicator weighting methods, this paper adopts Criteria Importance Through Intercriteria Correlation to remeasure the performance of corporate green transformation (Column 2). It measures the strength of comparison by calculating the standard deviation of indicators and uses the correlation coefficient matrix to capture the conflict between indicators, effectively solving the weight distortion problem caused by the entropy method neglecting the correlation between indicators (Diakoulaki et al., 1995). We control for more stringent year × industry × city high-dimensional joint FE to exclude unobservable macro- and meso-level shocks (Column 3). We exclude the sample during the COVID-19 pandemic from 2020 to 2023 (Column 4). To mitigate the problem of unbalanced covariate distribution, we use entropy balancing for weighted regression (Column 5). We include the independent variable lagged by one period to mitigate concerns about reverse causality (Column 6). We also add a cubic term to the independent variable. The second extreme point (approximately 1.082) that causes the curve shape to reverse is far beyond the maximum value of the independent variable, 0.844. The signs and significance of the lower-order terms remain consistent with the baseline results, while the additional turning point lies outside the observed sample range.

Although this study follows previous research (Fang, 2024; Li et al., 2024b) and constructs the corporate climate risk exposure measure using a text-based approach, some climate-related keywords may overlap with firms’ environmental practices. Therefore, we further conduct robustness tests using alternative climate risk measures. Specifically, we first decompose the baseline climate risk measure into transition risk and physical risk based on different types of climate risks, allowing us to examine whether different categories of climate risks generate consistent effects on corporate green transformation. Second, we remove keywords that are directly related to green transformation activities from the original indicator system and reconstruct an adjusted climate risk exposure measure. Table 6 reports the regression results using these alternative measures. The results show that transition risk (Column 1), physical risk (Column 2) and the adjusted measure excluding green-related overlapping keywords (Column 3) all exhibit significant inverted U-shaped relationships with corporate green transformation, indicating that the baseline findings remain robust.

To explore the pathways through which climate risks influence corporate green transformation, this paper introduces mechanism variables across three dimensions: TMT cognitive filtering, the external market environment and internal capital attributes. These mechanisms are not mutually exclusive. Instead, they capture different dimensions of corporate responses to climate risk: managerial cognition reflects how firms interpret and process external climate signals, market competition represents external pressure and incentive conditions and patient capital captures internal resource constraints affecting long-term transformation decisions. First, we use TMTs’ perceived supply chain uncertainty (SCUP) to capture how decision-makers interpret and respond to environmental fluctuations. This indicator is measured by the co-occurrence frequency of supply chain keywords and risk terms in annual reports (Benguria et al., 2022). Second, this paper uses market competition (COMP) to consider the effect of external competitive pressure on transformation motivation, measured as one minus the Herfindahl–Hirschman Index (1 − HHI) (Aghion et al., 2005). Finally, patient capital (P_CAP) captures the role of long-term funding in firms’ risk-bearing capacity and is measured by the proportion of shares held by long-term investors (Harford et al., 2018). While threshold models are commonly used to identify structural breaks or threshold effects, they may reduce the effective sample size in unbalanced panels and introduce selection concerns. Therefore, this paper uses subsample regression based on the turning point of the benchmark regression (He and Qi, 2022). The results are shown in Tables 7 and 8.

In the low-risk range (Table 7, Promoting Effects), ClimateRisk is significantly positively associated with both SCUP and COMP. Table 8 further shows that SCUP and COMP are significantly positively associated with GTrans. Taken together, these findings are consistent with the view that managerial perceptions of supply chain uncertainty and market competition are more prominent channels in the low-risk range. These results support H2 and H3. This result is consistent with Sharma (2000) and Aghion et al. (2005), who suggest that moderate external environmental pressures and market competition can enhance managerial risk perception and stimulate innovation, thereby promoting corporate green transformation.

In the high-risk range (Table 7, Inhibiting Effects), ClimateRisk is significantly negatively associated with patient capital; its association with SCUP is only marginally significant at the 10% level, whereas its association with COMP is insignificant. Table 8 shows that SCUP is not significantly associated with GTrans, while P_CAP is significantly positively associated with GTrans. Taken together, these findings are consistent with patient-capital contraction becoming a more prominent channel in the high-risk range. H4 is supported. This suggests that after the risk crosses the turning point, the capital-constraint mechanism becomes relatively more prominent and ultimately dominant, while the cognitive and competitive incentive mechanisms weaken. This finding is largely consistent with Bushee (1998) and Harford et al. (2018), who suggest that when uncertainty and risk levels are too high, firms’ willingness to invest in the long term is suppressed and the weakening of patient capital further undermines the long-term financial support needed for green transformation.

Column (1) of Table 9 reports the interaction effect between climate risk and corporate digitalization level. Corporate digitalization level (Digital) is measured by the proportion of digital transformation-related items in the year-end breakdown of intangible assets reported by listed companies (Zhang et al., 2021). The results show that the coefficient of ClimateRisk × Digital is significantly positive, indicating that the higher the level of digitalization, the stronger the promoting effect of climate risk on green transformation. The coefficient of ClimateRisk² × Digital is not significant. As shown in the left figure of Figure 5, as the level of digitalization increases from low to high, the turning point of the inverted U-shaped curve shifts significantly to the right. The turning point moves from approximately 0.330–0.390, a shift of approximately 18%. This indicates that digitalization enables companies to maintain an upward trend in green transformation even under higher-intensity climate risk shocks, thereby strengthening the incentive effect of climate risk.

Column (2) reports the interaction effect between climate risk and corporate financing constraints. This paper uses the Kaplan–Zingales Index (KZ Index) as a proxy variable for corporate financing constraints (Kaplan and Zingales, 1997). This index comprehensively examines financial indicators such as cash flow, dividend payments, cash holdings, debt-to-equity ratio and Tobin’s Q. A higher KZ value indicates a more severe limitation on external financing for the firm. The results show that the coefficient of ClimateRisk × KZ is significantly negative, while the coefficient of ClimateRisk² × KZ is significantly positive, indicating that financing constraints significantly alter the nonlinear relationship between climate risk and green transformation. Figure 5 (right side) shows that as the KZ level increases, the turning point of the curve shifts left from approximately 0.364–0.316, indicating that firms enter the risk-inhibiting green transformation range earlier. The curve generally flattens out, meaning that the firm’s response is weaker in both the incentive effect at low risk and the inhibitory effect at high risk. Financing constraints reduce the marginal driving effect of climate risk on green transformation and weaken its promoting effect.

The data for regional CPU are obtained from the Global Climate Risk Integration Database. This index uses deep learning algorithms to analyze news coverage from six major Chinese newspapers (Ma et al., 2023) and has been widely used in recent studies (Gao, 2025; Zhao et al., 2025). Provinces are classified as high-CPU if their annual uncertainty index exceeds the median, and low-CPU otherwise. To examine structural heterogeneity, we conduct Fisher’s permutation test (1,000 iterations) to compare coefficients across groups. The results in Columns 1 and 2 of Table 10 show that the differences in both the linear and quadratic terms of climate risk are significant at the 1% level. The larger absolute coefficients in the low-CPU subsample indicate stronger marginal sensitivity to climate risk. Climate risk exerts a stronger catalytic effect on green transformation in low-CPU regions at lower risk levels, but generates a more pronounced inhibitory effect once the risk exceeds a certain threshold, resulting in a steeper inverted U-shaped relationship. In regions with lower policy uncertainty, the relatively stable regulatory environment reduces firms’ wait-and-see costs and allows clearer expectations regarding environmental regulation (Trinks, Ibikunle, et al., 2022), thereby strengthening the initial innovation response to climate signals. However, this stronger policy alignment also reduces strategic flexibility when climate risks escalate. As compliance costs and capital risk premiums rise, firms in low-CPU regions may face sharper adjustment pressures, leading to a stronger inhibitory effect than firms in high-CPU regions that have already internalized uncertainty in their risk management strategies.

To further explore heterogeneity across industries, we divide the sample into labor-intensive and nonlabor-intensive sectors. Industry factor intensity is an important determinant of firms’ responses to environmental shocks (Zivin and Neidell, 2012). Following the 2012 China Securities Regulatory Commission Industry Classification Standard, firms are grouped into labor-intensive and nonlabor-intensive industries (Lu and Dang, 2014). Columns 3 and 4 show that the coefficient of ClimateRisk is significantly positive in both subsamples, indicating that moderate climate risk generally promotes firms’ green transformation across industries. However, the coefficient is larger for nonlabor-intensive firms, suggesting that these firms exhibit stronger responsiveness to climate risk. This pattern likely reflects the higher asset specificity and technological barriers in capital- and technology-intensive industries, where green transformation typically requires substantial research and development investment and specialized capital (Walker, 2013). As climate risks intensify, rising compliance costs and capital risk premiums (Trinks, Mulder, et al., 2022) therefore generate a more pronounced adjustment in these sectors.

To examine the differences in corporate behavioral responses to climate pressure, this paper draws upon the analytical framework in institutional theory regarding the potential structural differentiation between symbolic behavior and substantive action (Meyer and Rowan, 1977; Bromley and Powell, 2012), dividing corporate green transformation into substantive green transformation and symbolic green disclosure. Substantive transformation is measured by objective indicators such as green innovation, while symbolic disclosure is based on a green expression intensity index constructed from corporate annual report texts (Loughran and McDonald, 2011; Wan et al., 2021). Symbolic green disclosure is constructed using a separate green-expression dictionary and is not mechanically derived from the climate risk exposure measure. The results in Table 11 show that substantive green transformation and climate risk exhibit a significant inverted U-shaped relationship, declining after reaching a turning point. Symbolic green disclosure and climate risk show a significant positive linear relationship. This indicates a divergence between symbolic green disclosure and substantive green transformation under high climate risk. This pattern is consistent with, but does not by itself prove, greenwashing or opportunistic disclosure. It may also reflect stronger regulatory pressure, increased stakeholder communication needs or firms’ efforts to maintain legitimacy when substantive transformation is costly. To enhance comparability, this paper standardizes the two types of indicators and uses the standardized difference method to identify the deviation between symbolic disclosure and substantive action (Marquis and Qian, 2014; Lyon and Montgomery, 2015). By calculating the critical intersection point of the two types of responses, their standardized scores intersect at a climate risk level of 0.376, as shown in Figure 6. When the risk is below this threshold, the level of substantive transformation by firms is generally higher than that of symbolic green disclosure. When the risk exceeds this level, symbolic green disclosure begins to systematically exceed substantive transformation, and the gap widens as the risk increases. This indicates that in high-risk scenarios, when the space for substantive adjustment is limited, firms are more likely to rely on disclosure-based responses to maintain external legitimacy.

The theoretical contributions of this paper are as follows. First, existing research mostly discusses the promoting or inhibiting effects of climate risk from a linear perspective (Fang, 2024; Jin and Gao, 2025), while this paper reveals an inverted U-shaped relationship between the two, complementing the Porter Hypothesis by incorporating the investment-deterring effects of high uncertainty. We extend real options theory to the field of green transformation decision-making (Bolton and Kacperczyk, 2021), providing a unified explanation for the coexistence of existing promoting and inhibiting conclusions.

Second, this paper expands the research framework of external shock transmission and corporate strategic adjustment mechanisms. Existing literature emphasizes that supply chain shocks and uncertainties affect corporate investment through production networks (Barrot and Sauvagnat, 2016; Carvalho et al., 2021), and may lead to investment delays (Pindyck, 2007; Bloom, 2009), but rarely distinguishes differences in shock intensity. This paper, from the perspective of climate risk, supplements the theoretical connotation of the shift in the dominant mechanism under different risk stages of uncertainty shocks, expanding the applicability of related theories in the field of green transformation. This paper incorporates the logic of competition mechanisms and capital reallocation into the climate risk analysis framework, addressing the insufficient attention paid to the dynamic transformation of mechanism structures and enriching the explanation of corporate green transformation under climate risk shocks.

Third, this paper supplements the explanation of digital capabilities and financing constraints as boundary conditions for the effects of climate risk. Existing research has rarely discussed the role of digital technology in the green transformation mechanism under climate risk scenarios. This paper finds that digital capabilities are associated with stronger innovation incentives under climate pressure and a delayed triggering of inhibitory mechanisms, thereby expanding the economic consequences of digital transformation. This paper shows that financing constraints are associated with differences in the direction of green transformation and the curve shape and turning point of the relationship between climate risk and green transformation. From the perspective of curve structure and threshold adjustment, it deepens the explanatory power of financing constraint theory in the context of climate risk, providing a supplement to understanding the differentiated transformation paths of enterprises under the same climate pressure.

Fourth, this paper expands the applicability of decoupling theory in climate risk scenarios. Existing research indicates that under strong legitimacy pressure, companies may substitute symbolic disclosure for substantive change (Meyer and Rowan, 1977), exploit information asymmetry to engage in greenwashing and that increased disclosure does not necessarily imply substantive environmental improvement (Lyon and Montgomery, 2015). This paper identifies a divergence between symbolic disclosure and substantive green investment under high climate risk, thereby providing empirical evidence for decoupling theory in the context of climate risk.

This study has practical implications for policymakers, regulators and corporate management. First, climate risk exhibits an inverted U-shaped association with green transformation, suggesting that policymakers should pay attention to different risk stages and support firms’ green transformation through targeted subsidies, green finance support and institutional stability design, preventing a broad contraction in corporate green investment during high-risk phases. Second, the role of digital transformation in moderating the negative association between high-climate risk and green transformation suggests that digital and intelligent integration may enhance firms’ environmental resilience. Governments can enhance enterprises’ adaptability to climate risks through digital infrastructure construction and digital technology application incentives. Third, this paper finds that firms facing high climate risk exhibit greater divergence between disclosure and substantive transformation, which places higher demands on regulatory practices. Regulatory agencies should build a consistency review mechanism based on text analysis and cross-validation of behavioral data to improve the authenticity and binding force of environmental information disclosure.

This paper has limitations. First, although this study uses instrumental-variable estimation and extensive robustness tests, causal interpretation still depends on identification assumptions, particularly the exclusion restrictions of the instruments. Future research could exploit quasi-natural experiments, regulatory shocks or more granular data to strengthen causal identification and examine firms’ dynamic adjustment paths. Second, because the text-based climate risk measure may capture both disclosed climate risk exposure and managerial attention rather than realized physical damage alone, future research could combine textual indicators with firm-level physical climate exposure data. Third, this paper is based on a sample of Chinese enterprises. Given the specific features of China’s institutional environment, industrial structure and policy implementation intensity, the external validity of the findings should be further examined across different institutional and capital-market contexts. Future research could conduct cross-border comparisons to assess the universality of the nonlinear mechanisms and structural transformations of climate risk.

This paper examines the nonlinear impact of climate risk on enterprises’ green transformation. The results show a significant inverted U-shaped relationship between climate risk and corporate green transformation. This indicates that moderate climate risk is positively associated with corporate green transformation, while high risk is negatively associated with it. In the low-risk phase, green investment incentives are strengthened by managerial perceptions of supply chain uncertainty and market competition, while in the high-risk phase, the contraction of patient capital supply and the rise in the value of waiting options are associated with weaker long-term green investment. Corporate digital capabilities are associated with stronger positive responses to climate risk and a delayed turning point, but cannot eliminate the inhibitory mechanism in the high-risk phase. Financing constraints bring the turning point forward and flatten the nonlinear relationship. In the high-climate-risk range, companies increase symbolic green disclosure while decreasing substantive green investment, exhibiting a growing divergence between symbolic disclosure and substantive green transformation.