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

301 NTU Management Review Vol. 30 No. 2 Aug. 2020 The present research is subject to limitations. First, although CSR reporting can be viewed as a self-selecting process, ESG data sources are gathered from direct communications with companies, annual reports, 10-Ks, proxies, 20-Fs, sustainability/ CSR reports, and so forth. As such, the current study is different from that by Dhaliwal et al . (2011), who use an indicator variable that equals 1 if a firm discloses a standalone CSR report. Hence, we hardly control for the incentives of CSR disclosure. Accordingly, this is a study of association, not causation, and it is therefore vulnerable to the possibility that unobserved factors may drive the reported findings. Second, Lubos Pástor’s website only provides the monthly liquidity factor data, such that we cannot use daily data to estimate the β L . However, when we use the prior five years’ monthly data to estimate the β L and re-run the CSR performance models, the primary results are qualitatively the same. Third, we do not exclude the possibility that risk might be a moderator variable or mediator variable in which CSR performance can reduce the cost of capital. As such, whether risk is a moderator variable or mediator variable needs to be further investigated.