Core Insights - The research indicates that state-controlled media in China often exhibit a significant "optimistic bias," which can distort the true value signals for investors [1][11] - Social media platforms, such as Dongfang Caifu, serve as effective "correctors" of this bias, allowing investors to access more accurate information about companies [1][11] Research Background - The study analyzes data from 2009 to 2016, comparing 3.84 million posts from social media with 3.7 million reports from traditional media, focusing on their reporting on listed companies [8] - The unique media environment in China, characterized by traditional media's constraints and social media's relative freedom, provides a representative context for this research [8] Research Findings - Traditional media shows a significantly optimistic tone with a mean score of 0.3704 (77.9% positive), while social media has a mean score of -0.2028 (only 13% positive), indicating a negative bias [9] - During periods of positive reporting from traditional media, the correlation with social media's tone decreases significantly, confirming social media's ability to reveal traditional media's optimistic bias [9] - The study found that after the 2015 stock market intervention, the optimistic bias in traditional media increased, while social media's influence remained stable [11] Research Methodology - The research utilized machine learning techniques to analyze over 3 million news articles and 31 million social media posts, achieving over 90% accuracy in sentiment classification [10] - The analysis included various corporate characteristics and stock market trends to assess the relationship between media tone and future market performance [10] Research Implications - The findings highlight the role of social media as a supervisory mechanism in a heavily regulated news environment, enhancing market information efficiency [12] - This research contributes to understanding the interaction between media bias and financial markets, particularly in emerging market contexts [12]