Nyakurukwa, Kingstone2022-01-052022-01-052021https://hdl.handle.net/10539/32586A research report submitted in fulfilment of the requirements for the degree of Master of Commerce (50% research) in Finance to the Faculty of Commerce, Law and Management, School of Economics and Finance, University of the Witwatersrand, Johannesburg, 2021Online stock forums have emerged as an essential investing platform where multiple users can share their opinions about financial markets. This study examines the association between tweet features (bullishness, message volume and investor agreement) and market features (stock returns, trading volume and volatility) using 158 South African companies and a dataset of firm-level Twitter and StockTwits messages extracted from Bloomberg for the period 1 January 2015 to 31 December 2019. Firstly, the results from the study show that there is a general contemporaneous association between tweet features and stock market features. Secondly, no monotonic relationship is found between the magnitude of tweet features and the magnitude of market features. The coefficients of tweet features are found to be stronger in absolute terms at the extreme quantiles of returns. The BDS tests confirm the nonlinear relationship between stock returns and tweet features which implies that the relationship should be modelled using nonlinear specifications. Thirdly, the study finds no evidence that past values of tweet features can predict forthcoming stock returns using daily data. However, analysis using weekly and monthly data shows that past values of tweet features contain useful content that can predict the future values of stock returns. This implies that policymakers must implement appropriate regulations to deter the development of bubbles or crashes during the cycles of “greed” and “fear”. The lack of causal effects between tweet features and market features at the daily frequency implies that the JSE is dominated by institutional investors who are not driven by sentiment. The findings from the study corroborate the findings from previous studies which confirmed the dynamic nature of efficiency on the JSE.enThesis