Buy the rumors and sell the news - a sentiment analysis

Financial markets are broadly influenced by information conveyed across allmedia platforms. Previous research suggests that sentiment extracted fromfinancial news does affect the stock market (Tetlock, 2007), but to what extentdoes firm-specific sentiment affect a large-cap company that is highly integratedwith broad market movements? This study aims to answer and examine therelationship between sentiment, news volume and stock price volatility usingApple Inc. as a case study.Weekly aggregated financial news data from 2021 to 2024 is quantified throughFinBERT to extract sentiment scores. While combining the sentiment score withstock price data and different range-based volatility measures, four regressionmodels are estimated. The models estimated incorporate the following variables:sentiment, news volume, non-linear transformations, interaction variables, trad-ing volume and wide market volatility.Across the estimated models a clear pattern is discovered. Broad marketconditions are the dominant driver for all models, more specifically, wide marketvolatility and trading volume. However, news volume shows a clear U-shapednon-linear relationship with volatility. Sentiment shows limited explanatorypower, but when transforming sentiment into an interaction variable with newsvolume, sentiment contributes significantly in explaining volatility across allmodels.The Johnson-Neyman analysis indicates that the effect of sentiment onvolatility is significant during periods of either unusually high news volume orunusually low news volume. This analysis remains consistent across all modelspecifications.Overall, the results obtained in this study suggest that firm-specific senti-ment only has a marginal effect on the volatility of the particular stock, whilethe market-wide conditions remain as the dominant driver.