This study selects listed companies in Shanghai and Shenzhen A-shares from 2011 to 2021 as the sample. Companies engaged in the following industries are defined as high carbon emission enterprises: oil and gas extraction, electricity, heat, and gas production and supply, metal products manufacturing, petroleum processing, coking and nuclear fuel processing, ferrous metal smelting and rolling, chemical raw materials and chemical products manufacturing, non-ferrous metal smelting and rolling, chemical fiber manufacturing, non-metallic mineral products manufacturing, housing construction, non-ferrous metal mining, civil engineering construction, metal products machinery and equipment repair, construction decoration and other construction industries, non-metallic mineral mining, paper and paper products manufacturing, and wood processing and wood, bamboo, rattan, palm, and straw products manufacturing. Other companies in addition to the above companies are classified as low carbon emission enterprises.

Based on this classification, the sample was further screened as follows: (1) exclusion of ST, SST, and *ST companies; (2) exclusion of the financial and real estate industries; (3) exclusion of samples with significant missing values. Ultimately, 24,956 observations were obtained in this study. To avoid the effect of extreme values, all continuous variables are winsorized at the one percent level. In this study, companies from carbon-intensive industries are designated as the experimental group, while the remaining companies serve as the control group. All data utilized in the analysis are sourced from the CSMAR database and the CNRDS database.

To empirically assess the impact of carbon risk on corporate bankruptcy pressure, this study employs a DID model, a methodology commonly utilized in research to examine policy effects (Drysdale and Hendricks, 2018; Dewaelheyns et al., 2023; Cheng et al., 2024). The setup of the model is shown in Equation 1 below: Insolvency i , t = β 0 + β 1 C a r b o n i * P o s t t + β 2 X i , t + μ i + λ t + ε i , t (1) where, I n s o l v e n c y i , t stands for the Z-score of firm i in year t; Carbon is a dummy variable that indicates whether the enterprise is a high-carbon enterprise, and Post is a binary variable indicating the year t of the signing of the Paris Agreement; X i , t includes a set of control variables; μ i refers to individual fixed effects; λ t represents year fixed effects; and ε i , t denotes stochastic disturbances affecting corporate bankruptcy pressure. β 1 is the most concerned estimated coefficient in this study, reflecting the influence of carbon risk on corporate bankruptcy pressure.

Table 2 shows descriptive statistics. The mean of the Z-score is 6.029, with a standard deviation of 7.526. The Z-score ranges from 0.128 to 46.992, which indicates that the bankruptcy pressure of the listed companies is highly different. Our Z-score values and those of Ji et al. (2022) are consistent. The mean value of Carbon is 0.234, which signifies that the experimental group accounts for 22.4% of the population, suggesting that most firms have low carbon risk. Summary statistics for other variables generally align with the patterns observed in the existing literature (Pang et al., 2023).

The implementation of the DID model allowed us to evaluate the impact of carbon risk on the bankruptcy pressure faced by businesses. Table 3 reports the empirical results. In column (1), only the variable Carbon × Post is taken into consideration. In column (2), control variables are introduced to assess the impacts of other factors. Irrespective of the inclusion of control variables, the coefficients associated with the interaction term (Carbon × Post) exhibit significant positive values at the 1% significance level. This suggests a substantial alleviation of bankruptcy pressure following the implementation of the Paris Agreement, thereby supporting Hypothesis 1a.

In terms of control variables, our results are consistent with existing empirical studies that find that Size, Dual, and ListAge can significantly increase the bankruptcy pressure of corporates (Cho et al., 2021; Qin et al., 2023; Dewaelheyns et al., 2023). We also find that ROA and Cashflow can significantly reduce the bankruptcy pressure of corporates, which is consistent with our expectation and existing research. For example, Aziz et al. (2021) observed that enterprises with good performance in ROA and Cashflow have high profitability and low bankruptcy risk. All the other control variables exhibited statistical significance across all specifications, affirming the appropriateness of our selection of control variables.

An important premise for the use of the DID method is that the experimental group and the control group should exhibit a parallel trend before the implementation of the policy; otherwise, the estimated results will be biased. Specifically, we created time indicator variables including pre_3, pre_2, pre_1, current, and post_1, post_2, post_3, post_4, post_5. These variables represent 3 years before the implementation of the Paris Agreement, 2 years before, 1 year before, the first year of implementation of the Paris Agreement, the second year, third year, fourth year, and fifth year, respectively. As shown in Figure 1, the middle point in each vertical line is the parameter estimate value, and the two ends are the confidence interval when the confidence level is 95%. Pre_1 as the base year was removed from the regression. The coefficients of the regression terms in the years before the policy shock are not significantly different from zero. The coefficients are significant after the policy was implemented. Hence, the results satisfy the parallel trend test well.

We conducted a series of robustness tests to ensure the robustness of the empirical results, which included adopting a PSM-DID method, replacing explanatory variables, changing the sample period, redefining high-carbon industries, and conducing placebo tests. Irrespective of the specific robustness test applied, all the empirical results consistently reinforce the main conclusion of this study.

The analysis in this study explores the impact of carbon risk on firm bankruptcy pressure under the Paris Agreement. This section is based on theoretical analysis and aims to reveal the mechanisms through which this impact occurs.

Previous research has found a significant positive relationship between carbon risk and corporate innovation, with environmentally related innovative technologies tending to increase as pollution control costs rise, especially within heavily polluting enterprises (Lanjouw and Mody, 1996; Luo et al., 2023). According to the “green transformation” hypothesis discussed earlier, carbon risk will prompt enterprises to attach importance to technological innovation, optimize management processes, improve production efficiency, obtain competitive advantages and other resources, and reduce pollutant emission (Li, 2014).

Due to the characteristics of carbon emission risks, they pose multiple threats within the commercial environment, and their impact may become pronounced due to constraints on corporate financing. A series of regulations enacted to achieve green transformation are believed to mitigate the information asymmetry problem faced by high-carbon-emitting enterprises in this context (Zhu and Zhang, 2012). With the gradual strengthening of carbon risk, investors are becoming increasingly concerned about a company’s environmental performance (Pástor et al., 2021). Therefore, to meet the expectations of the government and investors, high-carbon emission enterprises will likely become more actively involved in information disclosure and carbon emission management to win the favor of investors and government funds.

In the context outlined above, enterprises can reduce bankruptcy pressure in the following two ways. First, through green innovation, they can improve their production efficiency, using green competitive advantages to enhance market share and their business capacity. Second, by disclosing information about their carbon emissions, enterprises can actively obtain the support of government funds and investors, solve the problems of high financing costs, and reduce the pressure of bankruptcy. In other words, enterprises can reduce bankruptcy pressure by increasing green innovation to alleviate financing constraints.

Referring to Hadlock and Pierce (2010), this paper uses the SA index to measure the financing constraints faced by enterprises. The SA index, which is constructed solely with variables such as company size and company age, which do not significantly change over time, exhibits strong exogeneity. This makes it a suitable proxy for financing constraints as it can help mitigate endogeneity problems. For the measurement of green innovation, this paper follows Jia et al. (2023) in using the natural logarithm of the total number of green invention patents of a company as the proxy variable of green innovation.

Drawing inspiration from Jiang (2022), the Equation 4 is constructed for mechanism testing: M i , t = θ 0 +   θ 1 C a r b o n i * P o s t t + γ X i , t + u i + η t + ε i t (4) where, M i , t represents the mechanism variables measuring financing constraints and green innovation, while the remaining items are consistent with model (1).

The empirical results, displayed in column (1) of Table 5, reveal a notable negative coefficient for the SA index, which suggests that carbon risk can mitigate a company’s bankruptcy pressure by reducing its financing constraints. The above conclusion is also consistent with previous empirical evidence that carbon risk can mitigate corporates’ bankruptcy pressure by easing financing constraints and promoting green innovation (Stamolampros and Symitsi, 2022). The empirical results, as presented in Table 5 column (2), show a significant positive coefficient for green innovation, implying that carbon risk can reduce corporate bankruptcy pressure by promoting green innovation within companies. Comparatively, while studies like those by Porter (1996), Bai and Tian, (2020) suggest innovation offsets regulatory costs, our findings further this by demonstrating how specific types of green innovation contribute to financial stability.

In the process of addressing carbon risk, enterprises of different sizes exhibit starkly different responses (Siedschlag and Yan, 2021). Based on the study of Tian et al. (2020), we establish average-sized companies as the standard and categorize companies into two groups: large-scale enterprises and small-scale enterprises. In Table 6, column (3) represents the regression results for large-scale enterprises, and column (4) represents the regression results for small-scale enterprises. These columns show that carbon risk significantly reduces bankruptcy pressures for small-scale companies, but its impact on large-scale companies is not statistically significant. This may be because although large companies desire to innovate and their R&D investment increases with the size of the company when facing pressure from outside (Wakasugi and Koyata, 1997), the defects in large enterprises’ flexibility and information accessibility tend to lead to inefficient innovation activities. This makes enterprise size irrelevant or even negatively correlated with innovation (Rogers, 2004). For small businesses, innovation incentives are more flexible when the businesses are faced with carbon risk, and simple corporate structures allow for more effective collaboration while avoiding bureaucracy (Simonen and mccnn, 2008). Therefore, carbon risk more effectively alleviates corporates bankruptcy pressure among small-scale companies than among large-scale companies.

Corporate ownership is a crucial boundary that demarcates different groups of enterprises as firms with diverse ownership structures exhibit significant differences in their cognitive logic, resource endowments, and operational approaches. Consequently, when facing carbon risk, these enterprises may adopt varying strategies. This study delves into the impact of carbon risk on bankruptcy pressures for state-owned enterprises (SOEs) and non-SOEs as differentiated groups. In Table 6, column (1) presents the regression results for non-SOEs, and column (2) presents the regression results for SOEs. These results show that carbon risk significantly reduces bankruptcy pressures for non-SOEs, while its effect on SOE remains insignificant.

This observation can be attributed to several factors. One plausible explanation is that non-SOEs possess a higher degree of flexibility and innovativeness (Wang et al., 2023). Generally, SOEs will inevitably undertake more corporate social responsibility (Liu et al., 2021), while SOEs tend to exhibit lower responsiveness to corporate performance. Conversely, managers in non-SOEs generally face heightened market pressures (Bradshaw et al., 2019), resulting in increased career apprehensions regarding the possibility of corporates bankruptcy. Therefore, non-SOEs are more able to reduce their bankruptcy pressure through green innovation than SOEs are. Moreover, non-SOEs tend to prioritize cost control and efficiency enhancement due to heightened market competition (Tang and Li, 2013). This focus on lean operations drives them to seek energy-efficient and emission-reducing solutions when faced with carbon risk, consequently mitigating carbon-related costs. In summary, the superior adaptive capacity of non-SOEs to mitigate bankruptcy pressures stemming from carbon risk is primarily attributed to their agility and innovation, as well as their emphasis on cost control and efficiency enhancement.

According to signal theory, intense market competition fosters adversarial relationships among companies (Muhmad et al., 2021). The more intense the market competition, the more competitors are eager to showcase their advantages through various means. Referring to Haushalter et al. (2007), this study adopts the Herfindahl-Hirschman Index (HHI) to measure the intensity of industrial competition. The specific calculation formula is H i j = ∑ X i j / ∑ X j 2 , where X i j represents the primary revenue generated by company i within industry j; and ∑ X j is the main business income of all enterprises within industry j. A smaller HHI index means more industry competition intensity. This study analyses the intensity of industry competition in relation to two levels, and divides it into the following two categories according to the median level of industry competition: high-level and low-level industry competition enterprises. In Table 6, column (5) represents the regression results for companies operating in highly competitive industries, and column (6) represents the regression results for companies operating in less competitive industries. In both columns (5) and (6), the coefficients are positive and significant, but the coefficient of enterprises in the industry with high competition intensity is larger, and the result is more obvious. There are several possible reasons for this phenomenon. Regarding the relationship between industry structure and firm characteristics, a highly competitive industry environment forces firms to develop a knowledge base that will enable them to pursue innovation and seize new market opportunities (Weerawardena et al., 2006). In addition, the intensity of industry competition also has a certain promotional effect on the diffusion of enterprise innovation results (Michalakelis et al., 2010). Both conditions make enterprises more inclined to actively respond to carbon risk and implement measures to address potential environmental and market challenges. Therefore, enterprises in highly competitive industries are more likely to proactively deal with the bankruptcy pressure brought about by carbon risk, meaning that carbon risk has a greater impact on enterprises in highly competitive industries.