Wisdom or Whims? Decoding Retail Strategies with Social Media and AI
A new study reveals retail investors on StockTwits use social media and artificial intelligence to shift trading strategies, often influenced more by social feedback like likes than by actual investment results. Analyzing 100 million messages with advanced AI, researchers found that positive posts about fundamental analysis predict higher stock returns, while bullish technical analysis posts signal lower returns, especially among Robinhood users. Only strategies based on fundamental analysis provide useful market information, highlighting both smart and impulsive investor behavior.
What it examines
This study uses social media data and large language models to analyze how retail investors choose and switch trading strategies. By classifying millions of StockTwits messages, the paper explores how strategy adoption is influenced by news, past performance, and social feedback, aiming to better understand retail investor behavior.
What it concludes
The research shows that retail investors' strategy choices are dynamic and shaped by news, performance, and social feedback. Fundamental analysis posts predict positive returns, while technical analysis posts often lead to losses. Applications include improving investor education, market monitoring, and designing tools for safer retail trading and financial regulation.
Evidence objects
Retail investors on StockTwits rapidly shift trading strategies, influenced by news, past performance, and social feedback, with AI tools like GPT-4 Turbo and BERT analyzing 100 million messages to track these changes.
key_findings bullet 1 · key_findings · validation V0
Surprisingly, positive posts about fundamental analysis predict higher future stock returns, while bullish technical analysis posts signal lower returns, especially among Robinhood users known for herding behavior and less informative trades.
key_findings bullet 2 · key_findings · validation V0
Social feedback, such as receiving 'likes,' drives strategy changes more than actual investment performance, revealing that only fundamental analysis-linked order flows are truly informative for market efficiency, highlighting both wisdom and whims in retail trading.
key_findings bullet 3 · key_findings · validation V0
This paper uniquely applies a two-stage LLM approachGPT-4 Turbo and BERTto classify retail investor strategies from vast social media data, extracting nuanced financial concepts. Its scale, integration, and real-time behavioral finance analysis are novel, offering compelling insights into dynamic strategy adoption, sentiment, and order flow informativeness.
key_findings bullet 4 · key_findings · validation V0
Raw abstract and provenance
- … and fragmented, rendering traditional text analysis methods inadequate. To overcome … strategy-specific sentiments on social media are related to actual trading activities. …
Source row: 2171 · abstract type: snippet