4.1. Sample Characteristics
Panel A of
Table 1 reports the summary statistics of all firm-month observations. The number of observations is 6832. The mean and median value of monthly buy-and-hold returns are 0.4% and 0.5%, respectively, reflecting the balanced distribution of the returns.
SOC and
ESG both show some positive skewness with the mean of 136 and 134 and median of 122 and 129, respectively. On the other hand,
ENV shows negative skewness with the mean of 146 and median of 156, implying that the firms generally have relatively low
ENV scores.
GOV displays a relatively balanced distribution with the mean of 124 and median of 122. For the firm size, we take a log value of the firm’s market capitalization, which results in a balanced distribution with the mean and median of 14.
Rating also shows a balanced distribution with the mean and median of 22.
Panel B of
Table 1 shows the correlation matrix among all variables. The correlation coefficients among
ESG and
ENV,
SOC, and
GOV are relatively high as can be reasonably assumed.
log_size is highly correlated with
rating (0.696) as rating agencies allocate a significant weight to the firm size when evaluating the credit quality of firms. The correlation coefficients among
rating and
ESG (0.355),
ENV (0.173) and
SOC (0.377) are relatively low, suggesting that credit ratings do not effectively reflect the aspects of ESG.
GOV, whose impact has been vastly examined in the literature in the context of corporate governance, shows the highest correlation with
rating (0.461) amongst others.
4.2. Impact of ESG on Bond Pricing
Table 2 reports the panel regression results of the impact of ESG on monthly bond returns from August 2010 to July 2015. The dependent variable is the monthly bond returns from
t to
t + 1. Independent variables are the information observable at
t. Throughout all specifications, the effect from the firm size and credit quality are controlled for along with the industry fixed effect. All estimates are reported along with the standard errors that are double-clustered by firm and date, following the method of Petersen [
16]. Except in column (1), the coefficients of ESG are negative and statistically significant across all models. When the interaction term between size and ESG, denoted by size_ESG, is introduced in column (2), however, the coefficient of the variable is 0.035 and statistically significant at the 1% level. Specifically, a one-unit increase in
size_ESG results in 0.035% increase in the bond return. This suggests that higher ESG scores of smaller firms are related to lower bond returns. Furthermore, lower bond returns imply that the bonds are priced higher at issuance, ergo lower funding costs. We discuss the implications in detail in
Section 4.
In column (3) where the interaction term
rating_ESG is introduced, the estimated coefficient of the variable is 0.001 and statistically significant at the 10% level. However, when all variables are used simultaneously,
rating_ESG loses its statistical significance and
size_ESG remains statistically significant at the 5% level with the coefficient of 0.029. Thus, the implication remains consistent. The higher the ESG scores and the smaller the issuer, the lower the bond returns. Another interesting finding is that the effect of ESG clearly exists separate from
rating, meaning that ESG provides complementary information to credit ratings in assessing credit risks of corporate bonds. Therefore, the overall results in
Table 2 are summarized as that ESG scores affect the debt financing cost for especially the small firms as well as signals credit risks not addressed by credit ratings.
Table 3 reports panel regression results using
ENV,
SOC and
GOV scores as the main independent variables and monthly bond returns as the dependent variable during the sample period. The formula is the same as the one used in
Table 2, except for the main variables. In column (1), it shows that a one-point increase in the environmental score leads to 0.081% decrease in bond returns, suggesting that bonds with higher environmental scores experience lower returns. Contrary to this, bond returns are increased by 0.119% for a one-point increase in the governance score. This conforms to the view that bond investors react negatively to the management’s effort in strengthening corporate governance which mostly benefits equity holders [
17].
In the following specifications, only
ENV shows significant impact on bond pricing when considering the impact of each criterion in conjunction with the issuer size. Specifically, a one-unit increase in the interaction term of
size_ENV results in 0.012% increase in bond returns. When each criterion is interacted with credit quality, no variable shows a statistically significant coefficient. Thus, depending on the firm size, environmental scores affect bond returns, or in other words, the cost of funding for issuers. This confirms and specifies the results in
Table 2.
The following may provide a rationale for such results. From the entrepreneur’s perspective, ESG is costly and difficult to implement. Firms should commit a considerable amount of investment to secure relevant resources including ESG data and specialists before implementing ESG firm-wide.
For this reason, large companies can endure related costs while small companies struggle. Therefore, if small firms actively implement ESG measures, it signals their dedication towards preserving the long-term value of the firms even by paying ESG taxes, thereby the lower returns or lower cost of debt financing for the issuers.
Table 4 reports panel regression results for robustness of the findings in
Table 3. The dependent variable is the same while independent variables such as
ENV_GOV, ENV_SOV, SOC_GOV and
ENV_SOC_GOV are newly introduced. The purpose of this test is to confirm whether the previously examined relationship between
ENV, SOC or
GOV and bond returns holds even after controlling for the additional interaction terms that may alter the relationship by their own interacting effects.
While the results confirm that the effect of ENV remains the same, it yields another interesting finding that has not been addressed in previous literature. As the positive coefficients imply, ENV and SOC substitute each other, as do SOC and GOV, whereas ENV and GOV do not. However, when interacting altogether, ENV, SOC and GOV become complementary to one another, thereby making it beneficial for bond issuers to promote all three ESG aspects simultaneously. Alternatively, if a firm decides to promote only two of the three aspects of ESG, the pre-existing effect on bond returns is nulled while promoting all three aspects of ESG contributes to lowering its debt financing cost.
Additionally, we check the existence of meaningful differences across the sample that are divided into two groups based on size.
Table 5 reports the differences in the mean values of
rating,
ESG and each of
ENV,
SOV, and
GOV between small and big firms. The results indicate that there are meaningful differences in the mean values of all test variables except for
returns between small and big firms, confirming that large firms have significantly higher ESG scores and rating. Lastly,
Table 6 reports robustness using double sorting based on size and ESG. The sample is divided into high and low scores of each criterion of ESG or ESG as a whole and also divided into small and big based on size. The test results confirm that the firms with low environmental scores show a statistically significant difference in returns (t-value of 1.71) between small and big firms. Such a result confirms our previous findings that environmental scores is an important determinant in bond returns especially for small size firms. In other words, smaller firms can effectively lower the debt financing cost by emphasizing the environmental aspect in firm-wide ESG implementation. While other return differences are statistically insignificant, the signs are all consistent with our regression results and conjecture.