Disclosure of specific information in social responsibility reports and the cost of debt financing
Description
Using machine learning for text analysis on a sample of social responsibility reports from Chinese A-share listed companies between 2009 and 2021, we have constructed an indicator to measure the level of specificity in the disclosure of these reports. We then analyze the effect of this specificity in CSR reports on the cost of corporate debt financing. Our study finds that the specificity of disclosure in CSR reports can significantly reduce the cost of corporate debt financing. This finding remains robust across various robustness tests. Additionally, the impact of specific disclosure in CSR reports on reducing the cost of debt financing is even more pronounced under conditions of higher corporate risk, such as excessive leverage and intense product market competition. Furthermore, further analysis suggests that the disclosure of specific characteristics in CSR reports has a more significant impact on reducing the cost of debt financing when industry disclosure is homogeneous, the information disclosure environment of firms is more perfect, and the intensity of information dissemination is higher.
Files
Steps to reproduce
We used A-share listed companies that disclosed social responsibility reports from 2009 to 2021 (with disclosure dates ranging from 2010 to 2022) as our research sample. We crawled the CSR reports of these companies from the CNINFO website using Python software. CNINFO is a commonly used source of firm-level data for China (www.cninfo.com.cn). The initial sample underwent the following treatments: (1) samples were excluded if report format conversion failed due to issues such as scanned versions; (2) samples from the financial industry were excluded; (3) samples with missing financial data were deleted; (4) samples labeled as ST or *ST were deleted; and (5) to avoid the influence of extreme values, continuous variables were winsorized at the upper and lower 1% levels. Finally, we obtained 7,567 company-year observations. All financial data used in our study were sourced from the China Stock Market & Accounting Research Database (CSMAR).