Score Driven Model for Realized Volatility of Crypto Assets - Innovative strategies for Modelling Cryptocurrencies Market Dynamics

This thesis explores the application of score-driven models in the analysis of realized volatility, with a particular focus on the cryptocurrency market. Realized volatility serves as a crucial measure for understanding asset price fluctuations over time, enabling more accurate risk management and investment decision-making. Given the highly volatile and rapidly evolving nature of cryptocurrencies, traditional financial models often fall short in capturing their dynamics. This study leverages score-driven models, which provide a robust framework for dynamically updating volatility estimates based on real-time market data. The research is structured into three main sections. The first chapter introduces the concept of realized volatility and its significance in financial markets, highlighting key data characteristics such as market microstructure effects and long memory properties. The second chapter delves into the theoretical foundation of score-driven models, presenting their statistical framework and discussing relevant probability distributions such as the Generalized Beta of the Second Kind (GB2) and its exponential variant (EGB2). The final chapter focuses on the practical implementation of these models using R, demonstrating their effectiveness through a case study on Bitcoin and Ethereum. By combining theoretical insights with practical applications, this thesis contributes to the ongoing development of sophisticated analytical tools for financial market analysis. The findings highlight the potential of score-driven models in enhancing risk management strategies and improving the accuracy of volatility analyses in the crypto market.