| Abstract | This thesis represents a pivotal advancement in the realm of behavioural finance, seamlessly integrating both classical and state-of-the-art models. It navigates the performance and applicability of the Irrational Fractional Brownian Motion (IFBM) model, while also delving into the propagation of investor sentiment, emphasizing the indispensable role of hands-on experiences in understanding, applying, and refining complex financial models.
Financial markets, characterized by ’fat tails’ in price change distributions, often challenge traditional models such as the Geometric Brownian Motion (GBM). Addressing this, the research pivots towards the Irrational Fractional Brownian Motion Model (IFBM), a groundbreaking model initially proposed by (Dhesi and Ausloos, 2016) and further enriched in (Dhesi et al., 2019). This model, tailored to encapsulate the ’fat tail’ behaviour in asset returns, serves as the linchpin for the first chapter of this thesis.
Under the insightful guidance of Gurjeet Dhesi, a co-author of the IFBM model, we delved into its intricacies and practical applications. The first chapter aspires to evaluate the IFBM’s performance in real-world scenarios, enhancing its methodological robustness. To achieve this, a tailored algorithm was crafted for its rigorous testing, alongside the application of a modified Chi-square test for stability assessment. Furthermore, the deployment of Shannon’s entropy, from an information theory perspective, offers a nuanced understanding of the model. The S&P500 data is wielded as an empirical testing bed, reflecting real-world financial market dynamics. Upon confirming the model’s robustness, the IFBM is then applied to FTSE data during the tumultuous COVID-19 phase. This period, marked by extraordinary market oscillations, serves as an ideal backdrop to assess the IFBM’s capability in tracking extreme market shifts.
Transitioning to the second chapter, the focus shifts to the potentially influential realm of investor sentiment, seen as one of the many factors contributing to fat tails’ presence in return distributions. Building on insights from (Baker and Wurgler, 2007), we examine the potential impact of political speeches and daily briefings from 10 Downing Street during the COVID-19 crisis on market sentiment. Recognizing the profound market impact of such communications, the chapter seeks correlations between these briefings and market fluctuations.
Employing advanced Natural Language Processing (NLP) techniques, this chapter harnesses the power of the Bidirectional Encoder Representations from Transformers (BERT) algorithm (Devlin et al., 2018) to extract sentiment from governmental communications. By comparing the derived sentiment scores with stock market indices’ performance metrics, potential relationships between public communications and market trajectories are unveiled. This approach represents a melding of traditional finance theory with state-of-the-art machine learning techniques, offering a fresh lens through which the dynamics of market behaviour can be understood in the context of external communications.
In conclusion, this thesis provides an intricate examination of the IFBM model’s performance and the influence of investor sentiment, especially under crisis conditions. This exploration not only advances the discourse in behavioural finance but also underscores the pivotal role of sophisticated models in understanding and predicting market trajectories. |
|---|