This study analyzes how social media sentiment influences stock market behavior in the United States, focusing on whether positive and negative public opinions have symmetric or asymmetric effects on asset returns. Leveraging a large-scale dataset of approximately 50 million Twitter messages referencing 2,557 publicly listed U.S. companies, the research examines how collective sentiment expressed online translates into measurable financial impacts. By expanding the scope both in number of firms and data frequency, the study aims to capture market reactions with a level of granularity that reflects real-time information flows.

The analysis combines automated natural language processing techniques with econometric modeling to transform unstructured social media content into quantitative sentiment factors. These sentiment signals are incorporated into GARCH-family models to assess their impact on stock returns and volatility at an hourly frequency. Unlike traditional approaches that treat news effects as linear or instantaneous, the methodology explicitly accounts for time delays, persistence, and asymmetry between positive and negative information. This allows the study to isolate how different types of sentiment propagate from social networks into financial markets.

The findings provide strong evidence that negative sentiment exerts a significantly larger influence on stock performance than positive sentiment, even when positive messages are more prevalent in volume. In most analyzed cases, sentiment effects materialize within an hour, highlighting the speed at which information travels from social platforms to markets. These results have direct implications for investors, risk managers, and regulators, as they demonstrate that social media sentiment constitutes a measurable and systematic source of market risk and return, reinforcing the importance of incorporating behavioral and informational factors into modern financial analysis.

Download the full report