Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.785180
Title: The impact of social mood on stock markets
Author: Pinto Souza, Tharsis Tuani
ISNI:       0000 0004 7970 7227
Awarding Body: UCL (University College London)
Current Institution: University College London (University of London)
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Abstract:
Assuming social media as a proxy for human activity, behavior and opinion, we aim to test the extent to which financial dynamics can be explained by collective opinion extracted from social media. First, we present an analysis of Twitter sentiment extracted from U.S.-listed retail brands. We investigate whether there is a significan causal link between Twitter sentiment, and stock returns and volatility. The results suggest that social media is indeed a valuable source in the analysis of financial dynamics, sometimes carrying more prior information than mainstream news such as the Wall Street Journal and Dow Jones Newswires. Second, we provide empirical evidence that suggests social media and stock markets have a nonlinear causal relationship. By using information-theoretic measures to cope with possible nonlinear causal effects, we point out large differences in the results with respect to linear coupling. Our findings suggest that the significant causal relationship between social media and stock returns is purely nonlinear in most cases. Furthermore, social media dominates directional coupling with the stock market, an effect that is not observable within linear modeling. Finally, we propose a model that predicts future correlation structure, based on a mechanism of link formation by triadic closure, that combines information from social media and financial data in a multiplex structure. The results demonstrate that the proposed model can achieve up to 40% out-of-sample performance improvement, compared to a benchmark model that assumes that correlation structure is time invariant. Social media information leads to improved models for all settings tested, particularly in the long-term prediction of a financial market structure. Our findings indicate that social media sentiment dominates directional coupling with the stock market in the prediction of individual asset dynamics as well as the overall market structure.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.785180  DOI: Not available
Share: