University of Patras
Non-parametric volatility estimation of cryptocurrencies using high frequency data
Abstract
dc:descriptionIn recent years, the cryptocurrency market has gained increased interest among investors, academics, policy makers and regulators all over the world, aiming to understand the unique characteristics and dynamics of cryptocurrencies price formation. The highly volatile nature and jump behavior in cryptocurrency markets as compared to traditional currencies, highlights the importance of defining a good proxy of volatility; this, will allow us to also account for the discontinuous movements in cryptocurrency prices to accurate model and forecast volatility. The overall purpose of our research is to model and forecast the volatility of cryptocurrencies, considering its critical role in many applications such as decision making, risk management and hedging. To this end, we consider the Heterogeneous Autoregressive model of Realized Volatility (HAR-RV). We utilize realized variance estimated from high frequency cryptocurrency price data. Specifically, we propose a non-parametric approach to estimate realized volatility and subsequently decompose it into continuous path variation and discontinuous path variation, to further estimate realized volatility jumps. First, we assess the predictive value of transaction activity in the Bitcoin Blockchain network, in an attempt to forecast the realized volatility of Bitcoin returns. We further explore extended versions of the HAR-RV specification, that include upside and downside jumps measures (positive and negative realized volatility jumps). The second area of interest focuses on the interconnectedness in the cryptocurrency market. To this end, we measure spillovers among major cryptocurrencies in terms of market capitalization, namely Bitcoin, Ripple, Ethereum and Litecoin. Specifically, we study the interconnectedness in the cryptocurrency market. We employ the proposed HAR-RV framework and its various extensions with the inclusion of different jump variations. In a univariate level of analysis, we examine the significance of heterogeneity and discontinuity to predict realized volatility of cryptocurrencies based on the HAR-RV framework. Additionally, we explore the spillover transmission mechanism among cryptocurrencies realized volatility by examining the relevance of realized volatility jumps and covariances. For this reason, we perform a comparative spillover analysis at a multivariate level of analysis. Multivariate HAR models (MHAR) are further examined in two versions, with and without the inclusion of covariances.
Degree
thesis:*- Grantor dc:publisher
- University of Patras
- Year dc:date
- 2022
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Tantoula, Maria
- Τάντουλα, Μαρία
Subjects
dc:subject × 17- Μεταβλητότητα
- Κρυπτονομίσματα
- Μηχανική μάθηση
- Διάχυση μεταβλητότητας
- Πρόβλεψη μεταβλητότητας
- volatility modelling
- Cryptocurrencies
- Volatility spillovers
- Machine learning
- Οικονομικά και Επιχειρήσεις
- Κοινωνικές Επιστήμες
- Χρηματοοικονομικά
- Οικονομετρία
- Economics and Business
- Social Sciences
- Finance
- Econometrics
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/52440