Exploring the interactions between categorised news sentiment and stock markets

<p dir="ltr">This thesis uses daily news sentiment indices generated from the financial news to explore interactions between news sentiment and market return, volatility and liquidity. The daily news sentiment includes the overall news sentiment and categorised news sentiments constructed from the Alexandria Contextual Text Analytics (ACTA) Equity data, categorised by the theme in daily frequency. Market return, volatility, and liquidity from six stock indices of global importance from the USA and the UK are utilised to explore the interaction between sentiments and the markets.</p><p dir="ltr">This thesis verifies the significant interactions between news sentiment and market variables through the Granger causality test and transfer entropy, indicating that historical news sentiment can explain the market return, volatility and liquidity. The analysis provides more significant conclusions for categorised news sentiment, especially some categories, than the overall news sentiment. This thesis also finds that interactions between news sentiment and market variables could vary across market periods. These findings suggest that categorising news sentiment could provide further insights into the market dynamics.</p><p dir="ltr">In addition, this thesis constructs linear models and non-linear models trained by machine learning techniques for news sentiment and market variables. While historical return, volatility and liquidity have significant and primary contributions to explaining market variables, adding categorised news sentiment indices to models further improves the performance of historical information in explaining the current market return, volatility and liquidity during different market periods. The analysis finds that the categorised news sentiments help predict the market liquidity, while no significant conclusions could be drawn about their impact on predicting return and volatility. The thesis highlights the value of categorising news sentiment by themes, which can improve the analysis of their impact on the financial markets and explain how these sentiment variables impact important market variables. The thesis provides valuable insights for academic researchers and practitioners.</p>