In empirical analyses, although the existing literature confirms the positive correlation between carbon risk and cost of debt, discrepancies exist in the findings to some extent. Delis et al. [
27] examine the syndicated loan market and compare the loan rate between fossil fuel and non-fossil fuel firms with a global sample. They show that only after the adoption of the Paris Agreement have banks started to price the stranded fossil fuel reserves risk, leading to an increase in the cost of credit for fossil fuel firms. Relatedly, Ehlers et al. [
2] study whether banks price carbon risk across all sectors to reflect a broader phenomenon beyond a specific sector. They also find a significant carbon premium in the syndicated loan market after the adoption of the Paris Agreement, although it is relatively low, at approximately 3-4 basis points. Interestingly, they confirm that the premium exists only for direct carbon emissions (scope 1), suggesting that banks are less concerned about firms’ whole carbon footprint than their direct emissions. In summary, both studies show that banks did not start pricing carbon risk until the Paris Agreement. Relative to them, Kleimeier and Viehs [
28] find that carbon emissions, arising either from scope 1 or scope 2, had a significant positive effect on loan spreads before the adoption of the Paris Agreement. They also find that for opaque firms, voluntary disclosure can significantly lower bank loan spreads. However, for transparent firms, voluntary disclosure has no impact. This result is consistent with the findings that reduced information asymmetries can lead to a lower cost of capital (e.g., [
29]). The reasons for the differences in conclusions may be the differences between sample firms. Although both Ehlers et al. [
2] and Kleimeier and Viehs [
28] use firm-level carbon emissions as a proxy of carbon risk, Kleimeier and Viehs [
28] cover firms that respond to CDP, and Ehlers et al. [
2] cover sample firms in the Trucost database, which includes not only firms that disclose their carbon emissions but also firms whose emissions are estimated through an input-output model. In addition, the authors use different methodologies to calculate firms’ emissions levels. Kleimeier and Viehs [
28] use industry- and firm-size-adjusted carbon emissions, and Ehlers et al. [
2] measure carbon risk as annual carbon emissions over annual revenues. The differences in their results may also be driven by different sample periods. Specifically, Ehlers et al. [
2] use the years 2006 to 2015 to denote the pre-Paris Agreement period, and Kleimeier and Viehs [
28] cover the years 2007 to 2013. Quiet differently, Delis et al. [
27] classify high- and low-risk firms based on whether they have fossil fuel reserves.