Addressing the power of news in financial markets: Analysing stock returns with GARCH models

Abstract

Risk management remains a paramount concern within the investment industry. Despite a wealth of literature and papers dedicated to stock volatility estimation and forecasting analysis, many experts redirected their attention to the relationship between news and stock returns. This thesis comprises three essays investigating the impact of news (newspaper headlines) on stock return indices by applying GARCH-type models (GARCH, EGARCH, and MGARCH). The initial essay delves into the returns estimation and forecasting of the GARCH model using various distributions in a portfolio context. The subsequent essay introduces news sentiment as an additional variable in GARCH and EGARCH models to effectively explore the impact on stock return estimations. The final essay addresses the correlations between the geopolitical risk news and energy stock (renewable and non-renewable energy stock) return fluctuations, explicitly focusing on the Russian-Ukraine war period. Multivariate GARCH models are employed. Chapter 2 seeks to assess the accuracy of the GARCH (1, 1) model in estimating and predicting portfolio returns and conditional variance for long-term investment, featuring two distinct distributions (normal and students’ t distribution). Weekly data, beginning in June 2010 and ending in June 2020 for ten years, were abstracted within the BRICS market. The findings underscore the superiority of the standard distribution assumption over the Student's t-distribution with GARCH (1, 1) for estimating and predicting conditional volatility. Chapter 3 analyses news impact on company stock returns and focuses on information within diverse industries. It evaluates news intensity, news sentiments (positive and negative news sentiment), and the VIX index (Benchmark index of the broad U.S. stock market) across individual companies and portfolio returns within selected APEC countries. Chapter 3 uses daily stock price data spanning 2017-2022 to employ analysis based on plain GARCH (1, 1) and EGARCH (1, 1) models. Models incorporating VIX log returns and varied news types as supplementary variables reveal results that diverge across industries and countries yet consistently affirm a robust correlation between the VIX index, news indexes, particularly news intensities and stock returns. The outperformance of GARCH over EGARCH becomes evident, highlighting stocks' heightened susceptibility to negative news. Chapter 4 scrutinises the repercussions of geopolitical risk-related news on renewable and non-renewable energy stock returns, particularly during the Russian-Ukraine war period. This examination involves testing three renewable energy stock indexes alongside three indexes representative of non-renewable energy stocks on a global, European, and US scale. By collecting daily energy stock index data from 2022 to 2023, the chapter applies the MGARCH model to elucidate correlations among the stock return indexes and GPR news. The analysis shows a positive correlation between world-level, European-level and American energy stock returns. The chapter also finds that increased headlines concerning geopolitical risk correspond to heightened renewable energy prices and decreased non-renewable energy prices.