Repository logo
 

The Effects of Social Media Sentiment on Financial Markets: A High-Frequency Study

Files

Thesis(1.66 MB)

Date

2024

Authors

Supervisor

Fernandez-Perez, Adrian
Indriawan, Ivan

Item type

Thesis

Degree name

Doctor of Philosophy

Journal Title

Journal ISSN

Volume Title

Publisher

Auckland University of Technology

Abstract

Social media has become one of the main communication channels over the last decade. It has reformed how investors acquire and exchange news and become their leading information source in the digital era. Social media platforms allow for the rapid dissemination of news, opinions, and sentiment to a vast number of market participants in real time. In line with the landscape change, recent studies have emphasised the significance of social media sentiment effects on financial markets. However, the majority of research focuses on the predictability of the sentiment derived from social media or assesses its impact on a specific type of financial asset. The mechanisms by which social media affects prices and investors are still not clear. In addition, whether social media sentiment captures information or noise is debatable. In this thesis, I examine the role of social media sentiment in financial markets from a market microstructure perspective and study its influence on various aspects of market dynamics. I produce a social media sentiment index from millions of real-time Twitter (now called X) messages through textual analytics, and I demonstrate the price impact of social media sentiment and its effect on market informational efficiency at a high-frequency level. Furthermore, I explore the spillover effects between social media sentiment and market volatility across various financial assets by employing Refinitiv MarketPsych analytics sentiment indices. The analysis at intraday and daily granularity captures the nuances of real-time social media sentiment impacts on market dynamics, demonstrating the mechanism of social media sentiment influencing financial markets. Hence, this thesis aims to contribute to the extant literature on how social media sentiment affects and interacts with financial markets in a high-frequency context. The first study of the thesis examines the mechanism by which social media sentiment affects stock prices. I assess the impact of Twitter posts on stock returns at the minute level. I find that social media sentiment can affect stock prices via trades. Specifically, an increase in buyer- (seller-) initiated trades has a significantly positive (negative) price impact. The impact is stronger with an increase in the number of tweets and sentiment, and persists even after controlling for volatility, liquidity shock, and limit-order activity. Both bullish and bearish tweets amplify the impact of trades on returns. It shows that the effect of social media sentiment is transmitted to stock prices through trades. The impact of Twitter sentiment on prices causes a permanent price movement at intraday, indicating that Twitter sentiment contains information. The second study investigates the impact of social media sentiment on the informational efficiency of financial markets. I examine the relationship between the aggregated tone of Twitter posts, i.e., the sentiment index used in the first study, and two commonly used market efficiency measures in empirical studies: return autocorrelation and variance ratio. The findings reveal that higher social media sentiment leads to higher intraday return autocorrelation and variance ratio the following day, indicating a decrease in market informational efficiency. I account for various influential factors, employ different sentiment analysis approaches, and consider different intervals for sentiment construction, all of which consistently support this relationship. Moreover, I demonstrate that social media sentiment impacts informational efficiency through the occurrence of herding behaviours among traders, with higher sentiment leading to heightened herding activity the following day. This study supports the notion that social media sentiment contributes to a decline in the quality of the information environment, resulting in informationally inefficient equity prices the following day. The third study delves into the dynamics of spillover effects between social media sentiments and market-implied volatilities among stock, bond, foreign exchange, and commodity markets. I find that informational spillover comes mainly from volatility indices to sentiment indices, with stock market volatility (VIX) being the most significant net generator. Within each asset class, there is a stronger spillover from volatility to the sentiment, but a marginal effect for the opposite direction. The connectedness between sentiment and volatility increases in turbulent economic periods, such as the Global Financial Crisis, Brexit, the US-China trade war, and the COVID-19 pandemic. Moreover, sentiment indices can switch from being a net receiver to a net generator of shocks during turbulent periods. This study shows that social media repeats existing news media signals, but some investors interpret repeated signals as genuinely new information. Overall, this thesis sheds light on the interplay between social media sentiment and financial market dynamics. It shows the mechanisms underlying the influence of social media sentiment on financial markets within the context of high-frequency analysis, contributing to the fast-growing research on the impact of social media on financial markets. Hence, the above findings have important implications for investors and market officials seeking to understand and better regulate social media as an information dissemination channel in the fast-changing environment. It provides insights for investors on utilising social media sentiment in real-time investment strategy. This thesis also emphasises the importance of regulatory frameworks when it comes to social media activity for market quality and stability.

Description

Keywords

Source

DOI

Publisher's version

Rights statement

Permanent link

http://hdl.handle.net/10292/17506

Collections

Doctoral Theses