Corporate digital transformation and financing cost of green bond: evidence from China

Against the backdrop of global climate governance, achieving the “Dual Carbon” goals and fulfilling the Paris Agreement require firms to undergo a dual transformation—decarbonizing both the “financing end” and the “production end.” On the financing side, firms must mobilize low-cost, long-term capital, while on the production side, they need to integrate real-time data and intelligent decision-making to reduce carbon intensity. Green bonds serve as a core instrument of this capital lever, with global issuance reaching USD 671.7 billion in 2024 ^[1]. By channeling funds into renewable energy, energy efficiency, and pollution control, green bonds help reduce carbon emission intensity (Flammer, 2021), and superior environmental performance further lowers the capital costs of green projects (Ambec and Lanoie, 2008). Meanwhile, corporate digital transformation (CDT) acts as a technological lever that reshapes firms’ production functions through Artificial Intelligence (AI), Internet of Things (IoT), and blockchain. Real-time energy monitoring enabled by digitalization reduces the carbon footprint per unit of output (Sun et al., 2025), while blockchain-based ESG (Environmental, Social, and Governance) data infrastructures enhance transparency and mitigate information asymmetry in green projects (Li et al., 2024a). When these capital and technological levers converge within the same firm, green bond credit spreads should theoretically benefit from a “double discount”, reflecting both improved green cash flows and lower agency costs. However, this proposition lacks direct evidence from bond-level microdata, particularly in China. Clarifying how CDT maps onto green bond spreads is therefore crucial for understanding firms’ financing costs and the efficient allocation of capital toward emission-reduction projects under the “Dual Carbon” goals.

Existing research presents divergent views on the capital market effects of CDT. The positive strand argues that CDT reduces corporate debt financing costs by enhancing information transparency and earnings persistence (Liu and Dang, 2025). Moderate digital transformation can further lower agency costs and irrational decision-making, thereby improving corporate resilience (Wang et al., 2025), easing financing constraints by mitigating information asymmetry and transaction costs (He et al., 2024), increasing media and academic scrutiny that curbs corporate tax avoidance (Chen et al., 2024), improving information transparency and bond rating quality (Wu et al., 2025), and facilitating more efficient access to green finance, particularly in tourism firms (Ye and Miao, 2025). In contrast, the negative view emphasizes that the technological and human capital adjustments associated with CDT may heighten financial risk, especially in manufacturing firms (Zhou and Zhang, 2025). Deep digitalization may also suppress financial asset holdings and capitalization (Sui and Yao, 2023), crowd out trade credit through asset specificity and bank credit channels (Zhou and Li, 2023), and increase the likelihood of corporate fraud due to weakened internal control and external supervision (Lu et al., 2025). In general, these findings suggest that the financial consequences of CDT are highly context-dependent and mediated by specific transmission mechanisms. Yet, in the green bond market, where environmental and financial attributes intersect, these mechanisms remain largely unexplored.

Drawing on signaling theory and agency theory, this study develops a testable hypothesis linking CDT to green bond financing costs. From a signaling theory perspective, CDT improves firms’ ability to collect, process, and disclose environmental and project-level information, thereby enhancing the credibility, traceability, and verifiability of green signals. This reduces information asymmetry between issuers and investors and lowers the required risk premium. From an agency theory perspective, CDT strengthens internal control and monitoring systems, improves the traceability of green bond proceeds, and reduces managerial discretion over fund allocation and environmental reporting. These improvements mitigate agency concerns such as fund diversion and greenwashing. Therefore, CDT is expected to reduce the financing cost of green bonds by alleviating both information asymmetry and agency frictions.

As one of the world’s largest green bond markets, China simultaneously pursues the “Dual Carbon” goals and the “Digital China” strategy. This policy-level coupling of green finance and digital transformation underscores China’s pivotal role in mobilizing capital for low-carbon transition while generating strong exogenous institutional pressure. Since Shanghai Pudong Development Bank issued China’s first labeled green bond in 2016, green bonds have become a core component of the country’s green finance system, providing refinancing channels for issuers and new pathways for green development (Dou and Qi, 2019; Yuan et al., 2025). China’s green bond issuance has consistently ranked among the world’s largest, with labeled issuance in the domestic market increasing by 60% from 2016 to 2019 ^[2]. Between 2016 and 2024, total issuance expanded from RMB 201.8 billion to RMB 683.3 billion, corresponding to an average annual growth rate of 16.5% ^[3]. Meanwhile, Chinese regulators have promoted the internationalization of sovereign green bond frameworks, attracting global investors and strengthening cross-border collaboration in green finance. Simultaneously, CDT in China is advancing rapidly. By 2024, core digital economy sectors accounted for 10% of GDP, reflecting the deepening integration of digital and industrial innovation ^[4]. This unique institutional environment, characterized by clear decarbonization objectives, active green finance experimentation, and heterogeneous digital transformation trajectories, provides a valuable setting for examining how digitalization shapes corporate behavior and capital market outcomes, while offering insights for other emerging economies and contributing to a more inclusive sustainable finance framework.

The empirical results show that CDT significantly reduces the financing cost of green bonds. Based on firm-bond observations of Chinese A-share listed companies from 2016 to 2024, CDT is negatively associated with green bond credit spreads, and this relationship remains significant after controlling for firm-level, bond-level, macroeconomic, and fixed-effect factors. The finding is robust to alternative variable measurements, entropy balancing, and instrumental-variable estimation. Further analyses indicate that CDT lowers green bond spreads through two channels: improving environmental information disclosure quality and strengthening internal control quality. Sobel and bootstrap tests provide additional support for these partial mediation effects.

This study makes three primary contributions. First, it extends the literature on the economic consequences of CDT by examining its role in the green bond market, thereby establishing a novel theoretical link between corporate digital capabilities and sustainable finance. Existing studies mainly focus on CDT’s effects on internal operations or traditional capital markets, showing that it promotes technological innovation (Fang and Liu, 2024), improves resource allocation efficiency (Jiang and Li, 2024), enhances labor investment efficiency (Wang et al., 2024), and reshapes workforce structures (Dou et al., 2023). Although recent research has begun to explore CDT’s environmental implications, such as its effects on environmental performance (Song et al., 2024) and green innovation (Zheng and Zhang, 2023; Sun et al., 2025), it has not been systematically incorporated into the green finance framework. This study fills this gap by elucidating how CDT affects green bond financing costs.

Second, this research expands the determinants of green bond financing costs by introducing a firm-level strategic factor. Prior studies emphasize macro-level policy influences, including carbon trading pilots (Yuan et al., 2025), collateral-based monetary policies (Xue and Yun, 2024), and the Belt and Road Initiative (Jian et al., 2022), but they offer limited explanations for cross-firm heterogeneity under similar policy environments. While some work considers green certification (Hu et al., 2022) and ownership structure (Sheng et al., 2021), this study captures the marginal effect of firms’ dynamic digital transformation.

Third, this study provides important practical implications for policymakers and managers. At the policy level, it supports the development of digital-transformation-oriented green finance frameworks, such as linking firms’ digital maturity to green bond certification and evaluation. For firms, it quantifies the financing benefits of digitalization and offers an empirical basis for designing digital investment strategies in environmental monitoring, ESG data governance, and disclosure enhancement.

Green bonds have emerged as a central instrument of sustainable finance, designed to mobilize capital for projects with explicit environmental benefits, such as renewable energy, pollution control, climate adaptation, and low-carbon transition (Flammer, 2021). Unlike conventional bonds, green bonds are characterized by a “green promise”: issuers commit to allocating proceeds to eligible green projects and disclosing information on the use of proceeds and environmental outcomes (International Capital Market Association, 2021). This feature creates a distinct risk–return structure. While conventional bond investors mainly price default risk, liquidity risk, maturity structure, and issuer credit quality (Jiang et al., 2025), green bond investors must additionally assess the credibility of environmental claims, the integrity of proceeds allocation, and potential greenwashing risk (Hu et al., 2022).

Green bonds also function as a commitment device in the transition toward a low-carbon economy. Under the “green bonding hypothesis,” green bond issuance exposes firms to stronger external scrutiny, third-party verification, and specialized disclosure requirements, enabling issuers to signal credible environmental commitments to capital markets (Flammer, 2021). Prior evidence suggests that policy support, green certification, environmental disclosure, and carbon-market mechanisms affect green bond pricing by reducing information asymmetry and strengthening investors’ confidence in issuers’ environmental credibility (Hu et al., 2022; Li et al., 2020; Yuan et al., 2025). Therefore, the financing cost of green bonds warrants separate analysis from the broader bond market because green bond spreads reflect not only conventional credit risk, but also green-specific pricing frictions related to environmental credibility, use-of-proceeds integrity, post-issuance monitoring, and greenwashing concerns. This distinction makes the green bond market a theoretically meaningful setting for examining the capital-market consequences of CDT, as digital technologies may improve environmental data traceability, disclosure credibility, and internal monitoring over green proceeds.

CDT represents a systematic strategic shift whereby firms integrate technologies such as AI, Blockchain, and IoT to reshape their production functions and organizational logic (Jiang et al., 2025). The literature on the economic consequences of CDT has grown rapidly and can be organized into three main streams. First, in terms of operational optimization, CDT enhances resource allocation efficiency and total factor productivity by improving data-processing capabilities (Jiang and Li, 2024; Wang et al., 2024). Recent evidence further shows that digitalization strengthens corporate resilience and eases financing constraints by mitigating information asymmetry (He et al., 2024; Wang et al., 2025). Second, the innovation-oriented literature demonstrates that CDT promotes both digital and green technological innovation by enhancing internal control quality and R&D efficiency (Fang and Liu, 2024; Zheng and Zhang, 2023). Third, the environmental sustainability literature suggests that CDT improves ESG performance by providing more granular and transparent data for sustainability reporting (Liu et al., 2024). The closest study to this paper is Jiang et al. (2025), which documents that CDT reduces corporate bond credit spreads in China through lower default risk and reduced information asymmetry. However, green bonds differ from general corporate bonds because their pricing incorporates not only credit risk and information asymmetry but also environmental credibility, proceeds earmarking, and greenwashing risk. This study therefore extends Jiang et al. (2025) by examining whether CDT reduces green-specific pricing frictions in the green bond market.

The financing cost of green bonds, commonly measured by the yield spread over comparable risk-free benchmarks, is shaped by both conventional bond-pricing factors and green-specific attributes (Li et al., 2020). Prior research identifies three broad categories of determinants: bond-level characteristics, green-credibility mechanisms, and issuer or institutional factors. First, bond-level characteristics such as credit ratings, maturity, and liquidity remain fundamental (Hu et al., 2022). Second, the green label and the credibility of third-party certification are pivotal. Green bonds with authoritative verification exhibit lower financing costs by mitigating greenwashing concerns (Hu et al., 2022; Zirek and Unsal, 2023). Higher-quality environmental information disclosure is also associated with reduced spreads (Tang et al., 2023). Third, the issuer’s institutional and strategic environment matters significantly. In China, state-owned enterprises often benefit from lower financing costs due to implicit government guarantees (Jian et al., 2022), while firms in high-pollution industries may experience larger spread reductions if their green issuance signals a credible strategic pivot (Wang et al., 2020). Macro-level factors, such as carbon trading pilots and regional digital infrastructure, also play a moderating role (Yuan et al., 2025). Despite this growing literature, little is known about whether firms’ internal digital maturity affects green bond pricing. This omission is important because CDT may serve as a technological enabler of both environmental disclosure and governance over green proceeds. This study therefore introduces CDT as a firm-level strategic capability that may reshape green debt pricing in the digital economy era.

Signaling theory posits that in environments of information asymmetry, high-quality firms can distinguish themselves from peers by transmitting costly and credible signals to the market (Ross, 1977; Spence, 1973). In the green bond market, investors face not only conventional uncertainty about issuer quality and repayment capacity, but also environmental integrity risk. This risk concerns whether bond proceeds are genuinely allocated to eligible green projects and whether the reported environmental outcomes are credible (Flammer, 2021; Hu et al., 2022). Although issuers may disclose sustainability-related information, such disclosures can be descriptive, backward-looking, and difficult for investors to verify. As a result, investors may discount green claims and require a higher risk premium when the credibility of environmental information is uncertain (Li et al., 2020).

CDT functions as a signal authenticator that enhances the quality of environmental information disclosure. The integration of “ABCD” technologies (AI, Blockchain, Cloud Computing, and Big Data) enables firms to capture project-level environmental metrics, such as real-time carbon emissions and energy efficiency, with unprecedented granularity and frequency (Liu and Dang, 2025). This digital infrastructure constitutes a high signaling cost that is prohibitively expensive for greenwashers to mimic without substantive green operations, thereby making green signals more difficult to imitate or manipulate. By improving the transparency, verifiability, and timeliness of environmental disclosures, CDT reduces the information processing costs for bondholders and alleviates concerns regarding greenwashing. Consequently, investors demand a lower information risk premium, which manifests as a reduction in the credit spread of green bonds (Tang et al., 2023).

From the perspective of agency theory, the separation of ownership and control creates incentives for self-interested managers to engage in opportunistic behaviors that may erode creditor value (Jensen and Meckling, 1976). In the context of green bonds, this manifests as an eco-agency problem, where managers may divert earmarked green funds toward high-risk general operations or engage in symbolic compliance to secure the greenium without delivering environmental benefits (Peng and Xie, 2024). Because green bonds require strict earmarking of proceeds, they impose higher monitoring costs on creditors than conventional bonds (Dou and Qi, 2019).

CDT can mitigate these agency frictions by strengthening internal control quality and improving the traceability of corporate processes. Digitalized monitoring systems allow firms to generate more traceable records of transactions, fund allocation, project milestones, and environmental performance indicators (Li et al., 2024b). These systems reduce managerial discretion, improve process standardization, and make abnormal fund diversion or symbolic compliance easier to detect (Ajakwe et al., 2025). In this sense, CDT improves the monitoring of green bond proceeds and reduces the likelihood of opportunistic behavior (Huang et al., 2025). By enhancing internal control quality, CDT can also strengthen creditors’ confidence that green bond proceeds will be used for their intended purposes and that environmental commitments will be honored. As perceived moral hazard, fund-misuse risk, and greenwashing concerns decline, investors are likely to demand lower governance-related risk compensation (Liu and Dang, 2025). Therefore, CDT may reduce green bond financing costs by alleviating agency frictions associated with the use of green bond proceeds and environmental reporting (Zhao et al., 2023).

While Jiang et al. (2025) show that CDT reduces generic bond spreads by lowering default risk and information asymmetry, the green bond setting introduces additional pricing frictions related to environmental disclosure credibility, proceeds earmarking, post-issuance monitoring, and greenwashing risk. Green bonds are therefore more sensitive to non-financial information quality and project-level governance than conventional bonds. By serving as both a signal-enhancing mechanism for environmental disclosure and a governance-enhancing mechanism for internal control, CDT can address the specific informational and agency vulnerabilities embedded in green bond markets. Accordingly, this study proposes the following hypothesis:

The year 2016 is widely recognized as the inaugural year for the development of green finance in China, marked by the issuance of the country's first standardized green bond. Subsequently, a series of supportive policies issued by the People's Bank of China (PBOC), the Ministry of Finance (MoF), and the China Securities Regulatory Commission (CSRC) have facilitated the rapid expansion of China's green bond market. To ensure data availability and consistency, this study focuses on labeled green bonds issued by Chinese A-share listed companies from 2016 to 2024. Observations with missing key variables are excluded and all continuous variables are winsorized at the 1st and 99th percentiles to mitigate the influence of outliers. Data on green bonds and corporate financial information are sourced from the China Stock Market and Accounting Research Database (CSMAR) and the Chinese Research Data Services (CNRDS) platform. The final sample comprises 533 firm-bond observations.

In the context of the digital economy, corporate digital transformation represents a critical adaptation to contemporary demands, enabling sustained development and innovation in the evolving business environment. Following the methodology of Wu et al. (2021), this study identifies and filters keywords related to the application of key digital technologies—Artificial Intelligence, Blockchain, Cloud Computing, Big Data, and other digital technologies—from corporate annual reports (Table 1). This process forms the ABCD + featured keyword lexicon, covering multiple essential dimensions of corporate digital development. PDF versions of all A-share listed companies' annual reports are collected from the official websites of the Shanghai and Shenzhen Stock Exchanges for the period 2016–2024 using Python. Text is extracted from these PDFs using Java PDFBox, and corporate digital transformation keywords are identified based on the ABCD + lexicon. The explanatory variable, CDT, is measured as the natural logarithm of one plus the total frequency of keywords identified from the lexicon.

Following prior studies on green bond financing costs and credit spreads (Li et al., 2020; Hu et al., 2022; Yuan et al., 2025; Zhou and Gao, 2026), this study uses the credit spread of green bonds to measure their financing cost. Spread is defined as the difference between the at-issue yield of a corporate green bond and the yield of a Treasury bond with the same issuance year and comparable maturity. Since many fixed-rate green bonds in China are issued at par, the coupon rate is used as a proxy for the at-issue yield when the reported issuance yield is unavailable. Under par issuance, the coupon rate closely approximates the yield to maturity at issuance because no issuance premium or discount exists. A higher Spread indicates a higher financing cost for the issuer. When a Treasury bond yield with exactly the same maturity is unavailable, the risk-free benchmark is obtained through interpolation.

The financing cost of green bonds may also be affected by issuer characteristics, bond-specific features, and local macroeconomic conditions. Following prior studies on green bond pricing and corporate bond spreads (Flammer, 2021; Hu et al., 2022; Li et al., 2020; Yuan et al., 2025), this study includes a set of firm-level, bond-level, and macro-level control variables. At the firm level, the controls include leverage ratio (Leverage), firm size (Size), fixed asset turnover (Turnover), return on equity (ROE), and state ownership (SOE). At the bond level, the controls include redeemability (Redeemable), credit rating (Rating), issuance amount (Amount), and bond maturity (Maturity). At the macro level, this study controls for local economic conditions (GDP). Table 2 provides detailed definitions for all variables.

Table 2 reports the descriptive statistics for the main variables used in this study. The final sample consists of 533 firm-bond observations. The mean value of CDT is 2.269, with a standard deviation of 1.125, indicating substantial cross-sectional variation in firms’ digital transformation levels. The median value of CDT is 2.639, suggesting that a considerable proportion of green bond issuers have already engaged in digital transformation to varying degrees. The dependent variable, the financing cost of green bonds (Spread), has a mean of 0.875% and a median of 0.645%, with a standard deviation of 0.786%. The spread ranges from −0.659% to 4.521%, suggesting that some green bonds are issued at yields below comparable Treasury benchmarks, while others require a substantial positive risk premium.

The firm-level control variables show reasonable distributions. Leverage has a mean value of 0.726, indicating that green bond issuers in the sample generally have relatively high leverage. Size, measured as the natural logarithm of the bond issuer’s total assets, has a mean of 26.329, reflecting the large scale of listed firms issuing green bonds. Turnover, measured by fixed asset turnover, has a mean of 2.047 and a standard deviation of 2.819, suggesting considerable heterogeneity in operating efficiency across issuers. ROE has a mean value of 0.021, indicating moderate profitability. The mean value of SOE is 0.696, showing that state-owned enterprises account for a large proportion of green bond issuers in the sample, which is consistent with the policy-oriented nature of China’s green bond market.

Regarding bond-level characteristics, the mean value of Redeemable is 0.099, indicating that approximately 9.9% of the sampled green bonds contain redeemable clauses. The mean value of Rating is 3.486, suggesting that the sample bonds generally have high credit ratings, mostly between AA+ and AAA. Amount, measured as the bond issuance scale in billions of RMB, has a mean of 2.385 and a standard deviation of 5.194, reflecting substantial variation in issuance size. Maturity, measured as the natural logarithm of bond maturity in years, has a mean of 1.006 and a standard deviation of 1.079, indicating substantial variation in bond tenors across green bond issuances. Finally, GDP, measured as the GDP growth rate of the city where the issuer is located, has a mean of 4.856, capturing differences in local macroeconomic conditions across issuers.

Table 3 reports the Pearson correlation coefficients among the main variables. The correlation matrix provides preliminary evidence on the bivariate relationships before conducting multivariate regression analyses. As expected, CDT is significantly and negatively correlated with Spread at the 1% level, with a coefficient of −0.325. This provides initial evidence consistent with Hypothesis 1, suggesting that firms with higher levels of digital transformation tend to have lower green bond financing costs. Several control variables also exhibit expected correlations with Spread. For example, Leverage, Size, Rating, and Amount are significantly negatively associated with Spread, indicating that issuer characteristics and bond-specific features are related to green bond pricing. However, these pairwise correlations should be interpreted with caution because they do not control for other firm-level, bond-level, macroeconomic, or fixed-effect factors. Therefore, multivariate regression analyses are required to provide more rigorous evidence on the relationship between CDT and Spread. The correlations among the explanatory variables are generally within an acceptable range. Although some variables show moderate correlations, such as the correlation between Leverage and Size, the variance inflation factor (VIF) values reported in the last column are all well below conventional thresholds. The highest VIF is 2.856 for Size, indicating that multicollinearity is unlikely to materially affect the subsequent regression results.

Table 4 presents the baseline results of CDT on the financing cost of green bonds. Column (1) shows the result without control variables and fixed effects and column (2) considers control variables without applying fixed effects. Column (3) only controls for industry and year fixed effects. Column (4) accounts for all control variables and fixed effects. In the fully specified model, the coefficient of CDT is −0.283 and significant at the 1% level. Based on the four regression settings, the negative coefficient of CDT on the financing cost of green bond shows high robustness. These findings suggest that corporate digital transformation can lower the financing cost of green bonds, thereby validating Hypothesis 1.

Following Zhang et al. (2025), this study adopts the digital transformation index from the “China Listed Companies Digital Transformation Research Database” in the CSMAR database as an alternative measure of CDT (CDT_new). In line with Tang et al. (2023) and Yuan et al. (2025), the new bond spread (Spread_new) is calculated using the yield of China Development Bank (CDB) bonds with matched maturities as the benchmark ^[5]. Compared to treasury bonds of the same term, CDB bonds also carry minimal risk and benefit from sovereign-level credit endorsement and high liquidity, though they are not tax-exempt.

Columns (1)–(3) of Table 5 report the results based on alternative variable measurements. In Column (1), both the dependent variable and the key explanatory variable are replaced by their alternative measures. The coefficient of CDT_new is −0.019 and statistically significant at the 1% level, indicating that firms with higher levels of digital transformation continue to face lower green bond financing costs when both key variables are redefined. Column (2) uses Spread_new as the dependent variable while retaining the original CDT measure. The coefficient of CDT remains negative and significant at the 10% level, suggesting that the baseline conclusion is not driven by the choice of the risk-free benchmark. Column (3) retains the original Spread measure but replaces the explanatory variable with CDT_new. The coefficient of CDT_new is −0.015 and significant at the 1% level. Overall, these results show that the negative relationship between CDT and green bond financing costs remains robust under alternative measurements of both the dependent and independent variables.

To further address potential covariate imbalance, this study employs entropy balancing as a reweighting approach. Entropy balancing directly adjusts the weights of control observations to match the covariate moments of the treated group, thereby improving balance quality and reducing model dependence (Hainmueller, 2012). After reweighting, the weighted control group closely matches the treated group across the selected covariates. Column (4) of Table 5 reports the entropy-weighted regression result. The coefficient of CDT remains significantly negative (β = −0.187), indicating that the baseline conclusion is robust after accounting for observable covariate imbalance. This result provides further support for Hypothesis 1.

To address potential endogeneity concerns arising from unobserved confounders and possible reverse causality, a two-stage least squares (2SLS) estimation is employed with an instrumental variable. Following Zhou and Li (2023), the internet penetration rate at the prefecture-city level is used as an instrument. This variable satisfies the relevance condition, as local internet infrastructure directly affects firms’ technological foundation and implementation costs for digital transformation, while also meeting the exclusion restriction, since it is a regional infrastructure indicator unlikely to directly influence firm-level bond financing costs. Columns (5) and (6) of Table 5 report the instrumental variable estimation results. Column (5) presents the first-stage regression, where the coefficient of the instrumental variable is positive and statistically significant at the 1% level. This result confirms the relevance of the instrument. In addition, the Kleibergen-Paap rk LM statistic is 26.583, rejecting the null hypothesis of under-identification. The Cragg-Donald Wald F statistic is 49.095, which is well above the Stock-Yogo critical value of 16.380, indicating that weak-instrument concerns are unlikely to be severe. Column (6) presents the second-stage regression results. After accounting for potential endogeneity, the coefficient of CDT remains significantly negative. This finding is consistent with the baseline regression and further supports Hypothesis 1, indicating that CDT reduces the financing cost of green bonds after addressing potential endogeneity concerns.

To further examine the mechanisms through which CDT reduces green bond financing costs, this study investigates two channels derived from the theoretical framework: the information transparency channel and the governance and agency channel. Specifically, environmental information disclosure quality (Envir) is used to capture the information transparency channel (Fan et al., 2024), while internal control quality (Internal) is used to capture the governance and agency channel (Chen et al., 2025).

Columns (1) and (2) of Table 6 report the results for the information transparency channel. Column (1) shows that the coefficient of CDT is positive and statistically significant at the 1% level, indicating that firms with higher levels of digital transformation tend to have better environmental information disclosure quality. This finding is consistent with the argument that digital technologies enhance firms’ ability to collect, process, and disclose environmental and project-level information in a more timely, traceable, and verifiable manner (Asif et al., 2023). Column (2) further includes Envir in the spread regression. The coefficient of Envir is significantly negative, suggesting that higher-quality environmental information disclosure is associated with lower green bond financing costs. Meanwhile, the coefficient of CDT remains negative and significant, indicating a partial transmission effect. These results support the information transparency channel: CDT reduces green bond spreads partly by improving the credibility and transparency of environmental information disclosure. Columns (3) and (4) report the results for the governance and agency channel. Column (3) shows that CDT is positively associated with Internal at the 1% significance level, suggesting that digital transformation helps strengthen internal control quality. This is consistent with the view that digital technologies improve process standardization, internal monitoring, and the traceability of fund allocation and project implementation (Zhao et al., 2023). Column (4) shows that the coefficient of Internal is significantly negative, indicating that stronger internal control quality reduces green bond financing costs. The coefficient of CDT remains negative and significant after controlling for Internal, suggesting that internal control quality partially channels the effect of CDT on green bond spreads.

Overall, the results in Table 6 provide evidence consistent with both proposed mechanisms. CDT lowers the financing cost of green bonds not only by improving environmental information disclosure quality, but also by strengthening internal control quality. These findings support the theoretical prediction that digital transformation reduces green-specific information and governance risks priced by investors in the green bond market.

To further examine the statistical significance of the proposed channels, this study conducts Sobel and bootstrap mediation tests. The Sobel test assesses whether the indirect effect of CDT on green bond financing costs through each mediator is significantly different from zero, while the bootstrap test provides confidence intervals without relying on the normality assumption. Panel A of Table 7 reports the Sobel test results. For the information transparency channel, the indirect effect through environmental information disclosure quality (Envir) is −0.0223, with a Sobel z-statistic of −1.9605 and a p-value of 0.0499. For the governance and agency channel, the indirect effect through internal control quality (Internal) is −0.0218, with a Sobel z-statistic of −2.0092 and a p-value of 0.0445. Both indirect effects are significant at the 5% level. Panel B presents the bootstrap results based on 500 replications. The 95% confidence interval for the indirect effect through Envir is [−0.0455, −0.0001], while that for Internal is [−0.0480, −0.0006]. Since both intervals exclude zero, the bootstrap results further support the significance of the two mediation channels. The direct effects remain negative after including the mediators, indicating partial mediation. These results suggest that CDT reduces green bond financing costs partly by improving environmental information disclosure quality and strengthening internal control quality, thereby supporting the information transparency channel and the governance and agency channel.

Against the background of global decarbonization and the rapid development of sustainable finance, this study examines whether CDT reduces the financing cost of green bonds. Using firm-bond observations of Chinese A-share listed companies from 2016 to 2024, this paper shows that CDT is significantly and negatively associated with green bond credit spreads. This evidence suggests that digitally transformed firms can obtain green financing at lower costs.

The baseline finding remains robust after controlling for firm-level characteristics, bond-level features, local macroeconomic conditions, and industry and year fixed effects. It also holds under alternative measurements of CDT and green bond spreads, entropy balancing, and instrumental-variable estimation. These results provide consistent support for the view that CDT is an important firm-level determinant of green bond financing costs. The mechanism analysis further indicates that CDT affects green bond pricing through two channels. First, CDT improves environmental information disclosure quality, thereby reducing information asymmetry and enhancing the credibility of green claims. Second, CDT strengthens internal control quality, which improves the traceability of fund allocation and mitigates agency concerns related to green bond proceeds. Sobel and bootstrap tests provide additional evidence that these two channels partially mediate the relationship between CDT and green bond spreads.

This study contributes to the literature in three ways. First, it extends research on the economic consequences of CDT from internal operations and conventional capital markets to sustainable finance. Second, it adds to the green bond pricing literature by identifying digital transformation as a firm-level strategic capability that helps explain cross-sectional variation in green bond financing costs. Third, it clarifies why green bonds warrant separate examination from general corporate bonds: their pricing reflects not only credit risk, but also environmental credibility, use-of-proceeds integrity, post-issuance monitoring, and greenwashing concerns. The findings also offer practical implications. For policymakers, the results suggest that digital infrastructure can complement green finance regulation by improving the transparency and monitorability of green projects. For firms, digital transformation is not merely a technological upgrade, but also a financing-relevant capability that can reduce the cost of green capital. For investors, CDT may serve as an additional signal when evaluating the credibility of environmental disclosures and the governance quality of green bond issuers.

This study has several limitations. The text-based CDT measure captures firms’ digital orientation disclosed in annual reports, but may not fully reflect the depth or effectiveness of actual digital implementation. Future research could use more granular indicators, such as digital patents, digital capital investment, IT human capital, or project-level digital infrastructure. In addition, although this study uses entropy balancing and instrumental-variable estimation to mitigate endogeneity concerns, future work could further exploit quasi-natural experiments or project-level data. Finally, the evidence is based on China’s green bond market. Comparative studies across different institutional settings would help assess the broader generalizability of the findings.

In sum, this study shows that corporate digital transformation helps reduce the financing cost of green bonds by alleviating green-specific information and agency frictions. As digitalization and decarbonization increasingly converge, firms’ digital capabilities are likely to become an important factor shaping sustainable finance outcomes.