Measuring volatility in cryptocurrency markets
View/
Keywords
Cryptocurrencies ; Volatility ; GARCH Models ; EGARCH Models ; Stochastic volatility ; ARIMA-GARCH ; ARFIMA (Long Memory Models) ; Value-atRisk (VaR) ; Expected Shortfall (ES) ; Macroeconomic variables ; Risk ; Risk management ; ForecastingAbstract
This thesis investigates the volatility dynamics of cryptocurrency markets and develops predictive models to enhance the understanding of price behavior in this highly speculative environment. Using a dataset of four major cryptocurrencies Bitcoin (BTC), Ethereum (ETH), Litecoin (LITE), and Dogecoin (DOGE) the study applies advanced econometric techniques to capture key stylized facts such as volatility clustering, heavy tails, and asymmetric responses to shocks. The methodological framework integrates GARCH(1,1) and EGARCH(1,1) models with Generalized Error Distribution (GED) innovations, ARIMA-GARCH models for joint mean-variance dynamics, GARCH-X models incorporating macroeconomic variables, Bayesian Stochastic Volatility (SV) models, and ARFIMA models to account for long memory.
The empirical results reveal that volatility persistence is extremely high across all assets, with leverage effects present for BTC, DOGE, and LITE, while ETH exhibits near-symmetric volatility responses. Incorporating macroeconomic variables demonstrates that inflation, liquidity growth, and energy prices significantly influence volatility, particularly for BTC, highlighting the interconnectedness between cryptocurrency markets and global financial conditions. Rolling one-step-ahead forecasts show that BTC and LITE achieve the most accurate predictions, ETH performs moderately well, and DOGE remains challenging to model, with VaR backtesting indicating systematic overestimation of downside risk. Bayesian SV models outperform GARCH-type models in capturing abrupt volatility shifts and regime changes, while ARFIMA results confirm minimal long memory for BTC and LITE, mild persistence for ETH, and near-zero persistence for DOGE.
These findings underscore the importance of flexible, heavy-tailed, and asymmetric models for accurate volatility forecasting in cryptocurrency markets. The study contributes to academic literature by providing a comprehensive comparative analysis of volatility models and offers practical insights for investors, risk managers, and policymakers. Recommendations include the use of advanced volatility models for risk assessment, the integration of macroeconomic indicators into forecasting frameworks, and the adoption of tail-sensitive risk measures such as Expected Shortfall for highly speculative assets.
Postgraduate Studies Programme
Χρηματοοικονομική και Τραπεζική με ειδίκευση στη Χρηματοοικονομική και Τραπεζική ΔιοικητικήDepartment
Σχολή Χρηματοοικονομικής και Στατιστικής. Τμήμα Χρηματοοικονομικής και Τραπεζικής ΔιοικητικήςNumber of pages
86Language
English
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα
Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα