Corporate Climate Risk and Bond Credit Spreads - 2024 - Finance Research Letters 3

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 1

H. Cang and C.

Li Finance Research Letters 67 (2024) 105741

remaining maturity from the yield-to-maturity of the corporate bond. In instances where the yield-to-maturity of a government bond
for a particular year was not available, we employed interpolation to estimate the value. A larger bond credit spread indicates a higher
cost of debt for enterprises.

3.2.2. Key independent variable: corporate climate risk


Following Li et al. (2024), we constructed the climate risk index (ClimateRisk) using the textual data extracted from corporate
annual reports obtained from Juchao3. The data were preprocessed using the Jieba word segmentation tool. We employed a
dictionary-based approach and utilized text analysis techniques to identify words within the annual reports that are related to climate
risk. Subsequently, the climate risk index was derived by calculating the ratio of the total frequency of the climate risk-related words to
the total word count of the annual report, multiplied by 100. A higher index value indicates a greater climate risk faced by the cor­
poration. Since the annual reports were published after the financial year, we used the lagged value of the climate risk indicator as the
key independent variable.

3.3. Empirical model

To examine the relationship between corporate climate risk and bond credit spreads, we estimated the following model:
Spreadsi,t = β0 + β1 ClimateRiski,t− 1 + ΦXi,t + Industryi + Yeart + εi,t (5)
The vector of covariates X includes the firm-level and bond-level determinants of Spreads. Appendix A provides the definitions for
all variables used in baseline regression. The Industry and Year fixed effects are controlled. The error term is denoted as ε, and the
standard errors are clustered at firm level.
Table 1 reports the statistical data of the main variables, with a total sample size of 1,275. The value of corporate bond credit
spreads (Spreads) ranges from 0.002 to 5.480, with a mean of 2.085. The mean and standard deviation of the main explanatory variable
(ClimateRisk) are 0.210 and 0.181, respectively. There is a significant difference between the maximum and minimum values, indi­
cating that the degree of climate risk varies among the firms.

4. Empirical test results and analysis

4.1. Baseline regression


Table 2 reports the results of the benchmark regressions. In Column (1), it shows that a higher level of climate risk in firms leads to
higher bond credit spreads, supporting our main hypothesis H1. In columns (2) and (3), we include bond- and firm-specific control
variables; the coefficients of ClimateRisk remain positive and significant at the 5 % level. It implies that ClimateRisk has a significant
effect on corporate bond credit spreads, with every one-unit increase in corporate climate risk leading to a 0.594 % increase in firms’
bond credit spreads.

4.2. Robustness tests

4.2.1. Instrumental variable regression


To address potential endogeneity problems caused by reverse causality and omitted variables, we enhanced the robustness of our
analysis by employing a two-stage least squares instrumental variable (2SLS-IV) method. Following Kling et al. (2021), we selected the
natural logarithm of the number of days with unusually high rainfall in the corporate’s head office city as our instrumental variable. As
a climate event, unusual rainfall is exogenous and does not directly affect bond credit spreads, and therefore meets the requirements of
an instrumental variable.
The results of the first stage analysis are presented in Column (1) of Table 3, and illustrate a positive and statistically significant
relationship between the instrumental variable (IV) and corporate climate risk. The second-stage results are detailed in Column (2),
with the coefficient of ClimateRisk remaining significant, providing further evidence that corporate climate risk positively affects bond
credit spreads.

4.2.2. Propensity score matching


We used the propensity score matching (PSM) method to address self-selection bias. A dummy variable for climate risk was
generated, taking a value of 1 when a company’s climate risk exceeded the yearly average, and 0 otherwise. Sample matching was
conducted using nearest neighbor matching based on the control variables in the baseline regression. The regression result is presented
in Column (3) of Table 3, and the statistical significance of ClimateRisk confirms the robustness of the main results.

4.2.3. Alternative proxy for corporate climate risk


In addition, we constructed an alternate proxy for climate risk by analyzing the management discussion and analysis (MD&A)
sections of the annual reports. The MD&A section typically included an in-depth discussion by the management, which aids in forming

3
Juchao Information Network is the information disclosure platform of the Shenzhen Stock Exchange. The official website is: www.cninfo.com.
cn.

You might also like