Time-varying higher moments in Bitcoin

Area of Science:

• Quantitative Finance
• Econometrics
• Computational Finance

Background:

• Cryptocurrencies, like Bitcoin, are novel investments with significant volatility.
• Traditional financial models struggle with the asymmetric and extreme price changes characteristic of cryptocurrencies.

Purpose of the Study:

• To model time-varying higher-order moments (scale, skewness, kurtosis) of Bitcoin returns.
• To investigate the predictive performance of Generalized Autoregressive Score (GAS) models with non-traditional innovations distributions for Bitcoin.

Main Methods:

• Utilized a modeling framework with time-varying higher-order moments.
• Estimated a series of Generalized Autoregressive Score (GAS) models.
• Employed non-traditional innovations distributions to capture cryptocurrency return dynamics.
• Compared model predictive performance using a Value at Risk (VaR) loss function.

Main Results:

• The study successfully modeled the complex, time-varying nature of Bitcoin's return distribution.
• GAS models with specific innovations distributions demonstrated superior predictive accuracy for Value at Risk.

Conclusions:

• Advanced econometric models are crucial for understanding and managing cryptocurrency investment risks.
• The findings provide a more robust framework for assessing Bitcoin's volatility and potential future price movements.