摘要
本文基于2010年1月-2020年5月东方财富股吧1.88亿条散户帖,从三个维度识别散户积极主义(关注信息深度、广度以及信息传播),并以深度学习LSTM模型搭建自然语言处理模型对帖子情感分析。基于此,本文探究积极散户能否提高上市公司审计质量。研究发现:积极散户有助于提高公司审计质量。积极散户以相对普通散户的信息优势发挥影响,通过在散户群体中传播信息、情绪,将其对公司的态度观点反映给审计师,为审计师提供可参考和重点关注的信息,从而提高公司审计质量。在散户负面情绪强烈时,这一作用更强,此时热衷传播信息的散户的作用更显著。本文的发现为促进中小投资者保护及提高公司审计质量提供了明晰的政策含义。
Based on 188 million posts on guba.eastmoney.com,we identify three types of individual shareholder activism:individual investors who pay attention to the depth or breadth of information,and individual investors who post frequently.Then we analyze the sentiment of posts with a natural language processing model based on a deep learning LSTM model.Using firm-level data during 2010-2019,we examine the impact of individual shareholder activism on audit quality.We find that individual shareholder activism is helpful in improving audit quality.Based on their information advantages,active individual investors spread information about firms to auditors by influencing other individual investors,which helps to improve audit quality.When individual investors have stronger negative sentiment,individual shareholder activism is helpful in improving audit quality more effectively.Furthermore,individual investors who post frequently have a greater impact on improving audit quality.Our findings have policy implications for protecting individual investors and improving audit quality.
出处
《会计研究》
CSSCI
北大核心
2022年第10期157-178,共22页
Accounting Research
基金
国家社科基金重大项目“创新驱动发展战略下全面塑造发展新优势的路径研究(20ZDA010)”
国家自然科学基金面上项目“公司事件、信息优势与投资者交易”(71772178)的资助
关键词
散户积极主义
投资者情绪
自然语言处理
大数据
审计质量
Individual Shareholder Activism
Investor Sentiment
Natural Language Processing
Big Data
Audit Quality
