Social media sentiment, behavioral biases, and abnormal returns: evidence from Indonesian Non-Cyclical Consumer Stocks

Social media has become an important source of information for investors, while stock price anomalies continue to challenge market efficiency. However, evidence on the combined role of social media sentiment, overconfidence, and herding behavior in the Indonesian stock market remains limited. This study examines the associations of social media sentiment and behavioral biases, particularly overconfidence and herding, with abnormal returns (AR) of Indonesia’s non-cyclical consumer sector stocks listed on the Indonesia Stock Exchange (IDX) during 2021–2025. Through a quantitative approach, this study employs panel data gathered from 119 firms selected through purposive sampling, comprising daily abnormal returns, a social-media-based sentiment index, overconfidence and herding proxies, and firm size as a control variable. The proposed model is tested using fixed-effects panel regression selected through Chow and Hausman tests. The reported main model shows positive and statistically significant associations for social media sentiment and herding, while firm size is also significant while included as a control variable. In contrast, overconfidence shows no statistically significance results after the reported robust standard-error adjustment. An alternative treatment is provided which excludes no-comment days yields consistent positive directions for sentiment and herding, but reduces the sample from 129,457 to 3,955 firm-day observations. The findings indicate statistically significant associations between social media sentiment, the herding proxy, and abnormal returns; they do not establish causal effects or a ranking of economic importance.