Can a 280-character message explain stock returns? Evidence from South Africa

Abstract

Purpose: The authors examine the contemporaneous and causal association between tweet features (bullishness, message volume and investor agreement) and market features (stock returns, trading volume and volatility) using 140 South African companies and a dataset of firm-level Twitter messages extracted from Bloomberg for the period 1 January 2015 to 31 March 2020. Design/methodology/approach: Panel regressions with ticker fixed-effects are used to examine the contemporaneous link between tweet features and market features. To examine the link between the magnitude of tweet features and stock market features, the study uses quantile regression. Findings: No monotonic relationship is found between the magnitude of tweet features and the magnitude of market features. The authors find no evidence that past values of tweet features can predict forthcoming stock returns using daily data while weekly and monthly data shows that past values of tweet features contain useful information that can predict the future values of stock returns. Originality/value: The study is among the earlier to examine the association between textual sentiment from social media and market features in a South African context. The exploration of the relationship across the distribution of the stock market features gives new insights away from the traditional approaches which investigate the relationship at the mean.