臺大管理論叢 NTU Management Review VOL.30 NO.2

287 NTU Management Review Vol. 30 No. 2 Aug. 2020 dimensions (Kim et al . , 2012; Kim et al . , 2014). If we exclude the corporate governance dimension from the CSR activities, the conclusion would lead to CSR performance being negatively related to the cost of capital. In order to distinguish between good and bad performers on each CSR dimensions, Bouslah et al . (2013) consider strengths and concerns separately and find that both corporate governance concerns and strengths positively affect a firm’s risk. They attribute this inconsistent impact to the fact that market participants do not agree on the value of the strengths and their impacts on the moments of cash flows. In addition, appropriate cost of equity capital estimates are difficult to obtain, so prior research adopts indirect estimates of cost of equity capital (Botosan, 1997). Several studies suggest that there are measurement errors in estimating the cost of capital, which can lead to spurious inferences, including studies by Goolsbee (2000) and Wang (2012). These measurement errors are also possible explanations for this observed contradiction on the positive relationship between corporate governance performance and the cost of capital. 4.2.3 Models Controlled for Industry Effects In order to control for industry effects of biasing our results, we separate our sample into 9 industries and add industry dummy variables into the CSR performance models (Model 1 to Model 4) as control variables in Table 8 and Table 9. The results show that these analyses yield inferences qualitatively similar to those without industry effects controlled. In particular, the coefficients of CSR_STR and TOTAL_CSR_STR on R_HAT_ MARKET , and the coefficients of CSR_CON and TOTAL_CSR_CON on FRET1 turn to be significant. Especially when we re-estimate Model 3 for cost of capital regressions by industry, untabulated results show that the effects of GOVERNANCE on the cost of capital are still positive, with the exception of R_HAT_MARKET regression (-3.037, t = -2.23) (standardized value) when observations are those firms in Retail Trade industry, indicating that the positive effects of GOVERNANCE on the cost of capital are not driven by industry.