Do ESG negative emotions on social media improve ESG disclosure quality? A study based on machine learning methods

ESG information has become a focus of attention for stakeholders. The study takes Chinese A-share listed companies from 2011 to 2021 as the research sample and employs the Word2vec algorithm and BERT model from machine learning methods to measure ESG negative emotions on social media (ENESM) based on the number of negative posts on investor interaction platforms. It finds that ENESM can compel companies to improve their ESG disclosure quality (EDQ). Mechanism testing shows that ENESM lead to a decrease in analyst attention, an increase in financing costs, and a decrease in customer stability for companies, forcing them to improve the EDQ. The more long-term institutional investors hold shares and the lower internal control quality, the stronger the positive impact of ENESM on the EDQ. Further research has found that compared to E (environmental) information and S (social) information, the positive correlation between ENESM and G (governance) disclosure quality is stronger. Although ENESM lead to companies with medium EDQ turning ESG greenwashing, strong regulation can effectively constrain this behavior. Non-ESG negative emotions on social media also drives companies to improve EDQ. ENESM further leads to analyst forecast dispersion and error. Moreover, increased financing constraints and bankruptcy risk compel companies to invest more in ESG disclosure costs in response to ENESM.The research has expanded the relevant literature on social media and EDQ, providing useful references for companies to cope with negative public opinion and improve EDQ.