Weibo sentiments and stock return: A time-frequency view

Area of Science:

• Computational social science
• Financial econometrics
• Behavioral finance

Background:

• Understanding the interplay between public sentiment and financial markets is crucial for economic analysis.
• Social media platforms like Sina Weibo offer vast, real-time data on public opinion.

Purpose of the Study:

• To investigate the relationship between detailed social media sentiments and the Chinese stock market.
• To determine the predictive power of microblog sentiments for stock returns.

Main Methods:

• Machine learning for sentiment classification of Sina Weibo microblogs into five categories: anger, disgust, fear, joy, and sadness.
• Wavelet analysis to examine the time-varying and frequency-specific linkages between sentiments and stock returns.

Main Results:

• Positive correlations were found between detailed sentiments and stock returns, particularly since October 2014 at medium to high frequencies.
• Sadness sentiment demonstrated a stronger association with stock returns compared to other sentiments.
• Stock market performance predominantly led Weibo sentiments, rather than the reverse, in periods of significant linkage.
• Detailed sentiments, unlike simple polarity, offer richer insights into market dynamics, with the market positively impacting bullishness and agreement.

Conclusions:

• Granular sentiment analysis from social media provides valuable information for understanding stock market behavior.
• The stock market influences public sentiment, but specific sentiments like agreement can also lead market movements, suggesting improved market certainty with reduced disagreement.