The relationship between public sentiment on Twitter and share price performance of South African retail banks

Abstract

Social media growth in South Africa is exponential. South African society has embraced the concept and adopted social media as a part of their everyday lives. They feel empowered by its reach and connected to people with similar ideas and interests. This gives society the ability to have their sentiments and opinions heard and they feel a sense of community and self-worth. They now have the power to punish those that they feel have done them wrong with very little consequence. They can influence mainstream and non-mainstream media debate and have a material impact on the prevalent sentiment in the country. This influence over sentiment affects the share price performance of South African retail banks, among other. Recent research proves that share price performance can be predicted by establishing sentiment on social media. The theoretical framework crosses behavioural finance with linguistics. Behavioural finance illustrates the influence of public sentiment on investor sentiment and in turn, the influence of investor sentiment on share price performance. Methods from the area of linguistics are applied to establish public sentiment as expressed on social media. The purpose of the study is to assess the significance of the relationship between public sentiment, as expressed on South African Twitter and the share price performance of South African retail banks on the JSE. Social media sentiment data was collected from Twitter via the Twitter API. Polarity and emotional sentiment was deduced from the text using NLP techniques. The closing balances of the share price data for the top five South African retail banks (ABSA, Nedbank, Standard Bank, Capitec and FNB) was collected over the same period. The data was analysed using Pearson’s correlation coefficient to establish a correlation between the social media sentiment explanatory variables and the retail banking share price performance dependant variable. Multiple linear regression was applied to these variables to establish if sentiment is an accurate predictor of share price performance. iii The key findings from the study is that social media sentiment on Twitter has a correlation with share price performance for several scenarios, at a 95% confidence interval. Polarity sentiment has a positive correlation, where negative sentiment on Twitter predicts a decrease in retail banking share prices three days after its observation. The emotional mood state of confusion is an accurate predictor of the next day’s share price performance, where an increase in confusion predicts a decrease in the next days’ retail banking share price performance. Social media sentiment is an accurate representation of public sentiment and public sentiment has a significant impact on investor sentiment. Research shows that positive sentiment in the markets specifically result in a decline in retail banking deposits. Questions that management in the retail banking sector must ask revolve around how much should be invested in protecting the corporate brand against social media sentiment, and what opportunities there are to harness the real-time capabilities of social media to predict the markets.