Bitcoin Volatility Modeling with ARMA-EGARCH and Realized Measures

Summary

The document examines Bitcoin return distributions and volatility, then outlines a modeling workflow using ARMA for returns and EGARCH for conditional volatility. It calculates log returns from closing prices and discusses descriptive statistics, quantile plots, and volatility clustering. For a volatility target, it uses higher-frequency price observations to estimate realized volatility, and compares rolling forecasts with iterative forecasts. The author reports that Bitcoin returns have fat tails and volatility clusters, and says rolling forecasts performed better in sample while out-of-sample returns were difficult to predict.

For distributional assumptions, the document reports that a generalized error distribution fit the observed tails better than normal or Student t alternatives, and favors EGARCH for longer-term volatility forecasts and asymmetric effects. These are presented as findings from the described analysis, not as broadly validated results. The author notes bold modeling assumptions and a lack of formal consistency checks. The dataset and period are limited, and the article provides no robust evidence that these forecasts deliver profitable trades after costs.

Key ideas

• Bitcoin log returns show fat tails and volatility clustering in the cited analysis.
• The workflow pairs an ARMA return model with EGARCH conditional volatility estimates.
• Higher-frequency observations can be aggregated into realized volatility measures.
• The author reports better in-sample performance for rolling than iterative forecasts.
• Forecast accuracy and trading value remain uncertain because validation is limited.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.